Modern Approaches in Machine Learning and Cognitive Science: A Walkthrough: Volume 4 (Studies in Computational Intelligence, 1117) [4] 9783031430084, 9783031430091

Citation preview

Studies in Computational Intelligence 1117

Vinit Kumar Gunjan Jacek M. Zurada Ninni Singh   Editors

Modern Approaches in Machine Learning and Cognitive Science: A Walkthrough Volume 4

Studies in Computational Intelligence Volume 1117

Series Editor Janusz Kacprzyk, Polish Academy of Sciences, Warsaw, Poland

The series “Studies in Computational Intelligence” (SCI) publishes new developments and advances in the various areas of computational intelligence—quickly and with a high quality. The intent is to cover the theory, applications, and design methods of computational intelligence, as embedded in the fields of engineering, computer science, physics and life sciences, as well as the methodologies behind them. The series contains monographs, lecture notes and edited volumes in computational intelligence spanning the areas of neural networks, connectionist systems, genetic algorithms, evolutionary computation, artificial intelligence, cellular automata, self-organizing systems, soft computing, fuzzy systems, and hybrid intelligent systems. Of particular value to both the contributors and the readership are the short publication timeframe and the world-wide distribution, which enable both wide and rapid dissemination of research output. Indexed by SCOPUS, DBLP, WTI Frankfurt eG, zbMATH, SCImago. All books published in the series are submitted for consideration in Web of Science.

Vinit Kumar Gunjan · Jacek M. Zurada · Ninni Singh Editors

Modern Approaches in Machine Learning and Cognitive Science: A Walkthrough Volume 4

Editors Vinit Kumar Gunjan Department of Computer Science and Engineering CMR Institute of Technology Hyderabad, India

Jacek M. Zurada Electrical and Computer Engineering Louisville, KY, USA

Ninni Singh Department of Computer Science and Engineering CMR Institute of Technology Hyderabad, India

ISSN 1860-949X ISSN 1860-9503 (electronic) Studies in Computational Intelligence ISBN 978-3-031-43008-4 ISBN 978-3-031-43009-1 (eBook) https://doi.org/10.1007/978-3-031-43009-1 © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors, and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, expressed or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This Springer imprint is published by the registered company Springer Nature Switzerland AG The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland Paper in this product is recyclable.

Contents

Implementation of Improved High Speed SHA-256 Algorithm from RTL to GDSII Using Verilog HDL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B. Srikanth, J. V. R. Ravindra, G. A. E. Satish Kumar, and Fahimuddin Shaik

1

Effect of Social Networking Advertisements (SNAs) on Attitudes and Purchase Intention Towards Brand Products . . . . . . . . . . . . . . . . . . . . . Venkata Subbiah Potala and A. S. Sathish

19

Design of Optimal Waste Management System Using IOT and Machine Learning Technique in Educational Institutions . . . . . . . . . . L. Sivayamini, C. Venkatesh, and Fahimuddin Shaik

33

Detection of COVID-19 Based on Deep Learning Methods: A Critical Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . J. Chinna Babu, Bolla Madhusudhana Reddy, K. Swapna, P. Yamuna, K. Sandeep Kumar Reddy, and N. Sumanth Performance Evaluation of GA, HS, PSO Algorithms for Optimizing Area, Wirelength Using MCNC Architectures . . . . . . . . . . Shaik Karimullah, D. Vishnuvardhan, Vinit Kumar Gunjan, and Fahimuddin Shaik An Enhanced Woelfel Image Noise Filter . . . . . . . . . . . . . . . . . . . . . . . . . . . . K. Riyazuddin, Shaik Bajidvali, B. Abdul Raheem, Shaik Karimullah, N. Merrin Prasanna, and Pabbati Swathi MHD Convective Flow of Chemically Reacting Viscoelastic Fluid Through an Infinite Inclined Plate via Machine Learning . . . . . . . . . . . . . Poli Chandra Reddy, B. Hari Babu, P. V. Sanjeeva Kumar, and L. Rama Mohan Reddy

41

53

71

81

v

vi

Contents

Improved Stockwell Transform for Image Compression and Reconstruction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Padigala Prasanth Babu, T. Jayachandra Prasad, and K. Soundararajan

95

Facemask Detection Using Bounding Box Algortihm Under COVID-19 Circumstances . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103 M. Hanumanthu, Shaik Karimullah, M. Sravani, Fahimuddin Shaik, P. Shashank, Y. Sravani, and K. VamsiKrishna Accelerated Addition in Resistive Ram Array Using Parallel-Friendly Majority Gates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111 J. Chinna Babu, Y. Suresh, R. Sudha Rani, S. Yasmeen, K. Siva Rama Krishna Reddy, and K. Harshavardhan Optimization of Area and Wirelength Using Hybrid BPSO Algorithm in VLSI Floorplan and Placement for IC Design . . . . . . . . . . . . 121 Shaik Karimullah, D. Vishnuvardhan, Vinit Kumar Gunjan, and Fahimuddin Shaik PAPR and SER Performance Analysis of OFDMA and SCFDMA . . . . . . 131 G. Obulesu, Shaik Karimullah, Fahimuddin Shaik, M. Nanda Krishna, C. Pavan Kumar, G. Divyanjali, and S. Mohammad Anas Food Detection with Image Processing Using Convolutional Neural Network (CNN) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141 K. Sreenivasa Rao, Fahimuddin Shaik, Munaga Sai Krishna, Sompalli Bhavya, Pothalam Bharat Teja, and Saginala Jaleel Basha Google Appstore Data Classification Using ML Based Naïve’s Bayes Algorithm: A Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147 J. Chinna Babu, Y. Suresh, Ajmeera Kiran, A. Ramesh Babu, and C. Madana Kumar Reddy Improved Spectral Efficiency Using Vehicular Visible Light Communication with 16-Bit DCO in OFDM . . . . . . . . . . . . . . . . . . . . . . . . . 159 Shaik Karimullah, D. Vishnuvardhan, Vinit Kumar Gunjan, and Fahimuddin Shaik Modelling of Symmetrical 13 Level and Asymmetrical 31 Level Generalized Cascaded Multilevel Inverters . . . . . . . . . . . . . . . . . . . . . . . . . . . 169 Bolla Madhusudana Reddy, P. B. Chennaiah, and J. Chinnababu Improved Radix-4 Fast Fourier Transform Algorithm Used for Wireless Communication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193 J. Chinna Babu and K. Naveen Kumar Raju Methodologies in Steganography and Cryptography–Review . . . . . . . . . . 205 G. Krishna Murhty and T. Kanimozhi

Contents

vii

Study of Secure Data Transmission-Based Wavelets Using Steganography and Cryptography Techniques . . . . . . . . . . . . . . . . . . . . . . . 215 K. Ravindra Reddy and Vijayalakshmi P. A Review: Object Detection and Classification Using Side Scan Sonar Images via Deep Learning Techniques . . . . . . . . . . . . . . . . . . . . . . . . . 229 K. Sivachandra and R. Kumudham Analysis of High Performance Optical Networks Using Dense Wavelength-Division Multiplexing Application . . . . . . . . . . . . . . . . . . . . . . . 251 L. Bharathi, N. Sangeethapriya, J. Prasanth Kumar, and G. Sandeep Wireless Sensor Network to Improve Security Performance and Packet Delivery Ratio Using FCL-Boost Based Classification Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263 N. Sangeethapriya, L. Bharathi, S. Jagan Mohan Rao, and A. N. L. Harisha To Analyse the Impact of Integration of Wind and Solar Power Generation System for Uttarakhand, Haryana and Rajasthan: A Scope of Machine Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 281 Himanshu Giroh, Vipin Kumar, and Gurdiyal Singh VLSI Implementation of an 8051 Microcontroller Using VHDL and Re-Corrective Measure Using AI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 293 Tushar Vardhan Zen K-Mean Energy Efficient Optimal Cluster Based Routing Protocol in Vehicular Ad-Hoc Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 305 A. C. Pise and K. J. Karande Dandelion Algorithm for Optimal Location and Sizing of Battery Energy Storage Systems in Electrical Distribution Networks . . . . . . . . . . . 315 Rajesh Patil and Varaprasad Janamala A Survey of Internet of Things Frameworks for Crowd Management System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 325 Jyoti Ambadas Kendule and Kailash Karande Performance Analysis of Patient Centric EHR Through Hyperledger Fabric . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 333 Somnath Agatrao Zambare and Namdev M. Sawant

Implementation of Improved High Speed SHA-256 Algorithm from RTL to GDSII Using Verilog HDL B. Srikanth, J. V. R. Ravindra, G. A. E. Satish Kumar, and Fahimuddin Shaik

1 Introduction The National Institute of Standards and Technology (NIST) is a federal agency, and it publishes Secure Hash Algorithms, a member of the family of encrypted hash functions [1]. Federal Information Processing Standard (FIPS), inclusive SHA-0, SHA-1, SHA-2, and SHA-3. In which SHA-256 of SHA-2 family was designed by the National Security Agency (NSA) to serve as cryptographic Digital Signature Algorithm and is certified with FIPS PUB 180–4, CRYPTREC and NESSIE [2]. A cryptographic algorithm called SHA-256 converts data with variable lengths to data with fixed lengths [3]. The input is of variable length i.e.; “N” 512 blocks and each block is of size 512 bit.The “N” 512 bit input must be less than 264 bits [3]. The resultant length of SHA-256 is of 256 bit. This 256 bit is called as a hash or message digests [4]. The final hash acts as a digital signature in cryptographic security, and cryptocurrency. This paper proposes few optimization techniques in the architecture of SHA-256 hash algorithm namely addition of independent variables in the message compression function and addition of independent variables in the message scheduler function for improving hashing algorithm to minimize the critical path and to achieve less area utilization, power consumption and cells count that improves the performance of SHA-256 design [5].

B. Srikanth (B) · J. V. R. Ravindra · G. A. E. Satish Kumar Department of Electronics and Communication Engineering, Vardhaman College of Engineering, Hyderabad, India e-mail: [email protected] F. Shaik Department of Electronics and Communication Engineering, AITS, Rajampet, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 V. K. Gunjan et al. (eds.), Modern Approaches in Machine Learning and Cognitive Science: A Walkthrough, Studies in Computational Intelligence 1117, https://doi.org/10.1007/978-3-031-43009-1_1

1

2

B. Srikanth et al.

2 Normative SHA-256 System Architecture Initial hash values A[0], B[0], C[0], D[0], E[0], F[0], G[0], H[0] are given by NSTand each of size 32 bit SHA-256 constants K[0], K[1], …, K[63] are given by NST and each of size 32 bit as shown in Fig. 1. The SHA-256 consists of a main module and two sub modules like message scheduler & message compression [6]. The message scheduler performs the message scheduling operations and message compression performs the message compression operations [7]. These are discusssed in detail in Sects. 2.3 and 2.4 respectively. SHA-256 was published by NIST and they proposed a set of 8 initial hashes (words) A[0], B[0], C[0], D[0], E[0], F[0], G[0], H[0] and they proposed also proposed a set of 64 constants (words) K [0], K[1]…K [63]. SHA-256 uses sixty-four constant 32-bit words, K[0], K[1], K[2], …, K[63] as shown in Fig. 2. These words represent the first thirty-two bit words of the fractional parts of the cube roots of the first sixty-four prime numbers [8]. These words are constants in hex and are (left to right). SHA-256 needs an input of “N” 512 bits where “N” 512 < 264 as shown in Figs. 3 and 4. Let’s say the input passed is M bits. The algorithm needs to be pre-processed before it is passed as input to SHA-256 Algorithm [9]. Preprocessing includes three sub-steps which are the following. • The input message was padded, M • Blocking the message for parsing • A, B, C, D, E, F, G, and H are the initial hash values that are set. (Refer to Fig. 3) Fig. 1 SHA-256 block diagram. Note SHA-256 Input given here is paddea and parsed message of size 512 bit

Implementation of Improved High Speed SHA-256 Algorithm …

3

Fig. 2 Conventional SHA-256 main and sub-module block diagram

Fig. 3 SHA-256 Initial Hash values Note Each hash mentioned above is of size 32 hex bit

428a2f98 71374491 b5cofbcf e9b5dba5 3956c25b 59f111f1

923f82a4 ab1c5ed5

d807aa98 12835bo1 243185be 55oc7dc3 72be5d74 8odeb1fe

9dbco6a7 c19bf174

e49b69c1 efbe4786 ofc19dc6 240ca1cc 2de92c6f 4a7484aa 5cboa9dc 76f988da 983e5152 a831c66d boo327c8 bf597fc7

c6eoobf3 d5a79147 o6ca6351 14292967

27b70a85 2e1b2138 4d2c6dfc 5338od13 65oa7354 766aoabb 81c2c92e 92722c85 a2bfe8a1 a81a664b c24b8b7o c76c51a3 d192e819 d699o624 f4oe3585 1o6aao7o 19a4c116 1e376co8 2748774c 34bobcb5 391cocb3 4ed8aa4a 5b9cca4f 682e6ff3 748f82ee 78a5636f 84c87814 8cc7o2o8 9obefffa

a45o6ceb bef9a3f7 c67178f2

Fig. 4 SHA-256 Constants

2.1 Padding the Input Message (M) In order to ensure that the padded input message, M, is multiple of 512 bits with respect to SHA-256, padding is required. Assume that the message M is composed

4

B. Srikanth et al.

Fig. 5 Padding of input message

of l bits of “1” data up until the end of the message, followed by k bits of zero data, where k is the lowest, a positive solution to the equation l + 1 + k = 448 mod 512. Then add the 64-bit block that corresponds to the binary version of the number l expression [10]. For instance, the (8-bit ASCII) message “abc” is padded with one bit, 448 − (24 + 1) = 423 zero bits, and then the message length to create a 512-bit padded message because 8 * 3 = 24 is the message’s length [11]. Now, as illustrated in Fig. 5, the length of the padded message should be a multiple of 512 bits.

2.2 Parsing the Padded Input The padded message of the above step must be parsed into N 512-bit blocks. Now, the message after pre-processing stage is fed to Message Compression Function [12] as shown in Fig. 6. The three main tasks to be performed after preprocessing are the following. • Message Scheduler Function. • Message Compression Function. • Intermediate Hash Function.

Fig. 6 Parsing of padded message

Implementation of Improved High Speed SHA-256 Algorithm …

5

Fig. 7 Conventional generation of W[0] to W[15] Fig. 8 Conventional generation process of W[0] to W[15]

2.3 Message Scheduling Function Each padded input message block, M (1), M (2) … M (N) is sent to Message Scheduling Function [13]. In message scheduling algorithm we need schedules of W[0] to W[63]. Let the M (1) be the only pre-processed input of 512 bit and it is sent into W[0], W[1] … W[15] as shown in Fig. 7. Design now needs to generate W[16] to W[63]. This task is done by message scheduler [14] as shown in Figs. 8 and 9.

2.4 Mechanism of Message Compression The register contents were changed to implement the message compression function [15]. The eight initial hashes proposed by NIST A[O], B[O], C[O], D[O], E[O], F[O], H[O] (refer to Fig. 3) are stored in eight 32-bit shift registers A[ii], B[ii], C[ii], D[ii],

6

B. Srikanth et al.

Fig. 9 Conventional generation of W[16] to W[63] in message scheduler function

E[ii], F[ii], G[ii], H[ii] and compression function runs for 64 rounds as shown in Fig. 10, i.e.; ii variable runs for 64 times generating the following Figs. 11, 12, 13, 14, and 15. The generation of newA is the critical path in this core.

Fig. 10 Conventional SHA-256 message compressionblock diagram

Implementation of Improved High Speed SHA-256 Algorithm …

7

Fig. 11 SHA-256 Conventional message compression block diagram

Fig. 12 Generation of sumA and sumE in Normative message compression block diagram

2.5 Intermediate Hash Generation For variable, ii = 0; A[ii] turns to be A[0] which is our initial hash is passed to B[ii + 1] = B[1] = newB and in next round A[1] is passed to B[2]. Similar operation is done on newC, newD, newF, newG, newH as shown in Fig. 16.

8

B. Srikanth et al.

Fig. 13 Generation of ch and maj in normative message compression block diagram

Fig. 14 Generation of newE in normative message compression block diagram

In this way for 64 rounds, i.e.; 0 < = ii < = 63 we get A[1] to A[64], B[1] to B[64], …., H[0] to H[64]. Final hash = concatenation. (A[O] + A[64], B[O] + B[64], C[O] + C[64], D[O] + D[64], E[O] + E[64], F[O] + F[64], G[O] + G[64], H[O] + H[64]) as shown in Fig. 17.

Implementation of Improved High Speed SHA-256 Algorithm …

Fig. 15 Generation of newA in normative message compression block diagram

Fig. 16 Intermediate hash generation

9

10

B. Srikanth et al.

Fig. 17 Generation of final hash (Message Digest)

3 Improved SHA-256 System Architecture 3.1 Proposed Work 1. USAGE OF CARRY SAVE ADDER (CSA): The creation of newA variables for each round of the compression function is, as was previously said, the critical step in the SHA256 design. This involves modulo addition of 2 w (w = 32 bit) on seven variables, they are H[ii], k[ii], W[ii], sumA, maj, ch, sumE. Usage of Carry Save Adders, lessen the delay caused by the propagation time of carry by separating the generation routes of sum and the carry. CSA accept 3 operands as inputs and so, the working variable A can be computed using just 5 CSA as shown in Figs. 18, 19, 20, 21, and 22. So, in the design every modulo addition 2 w is now replaced with a carry save adder (CSA).

4 Reducing 3 Adders to 2 Adders in Message Scheduling In the Fig. 8, the addition of W[i-15] = q and W[i-14] = s are independent to each other and also to the message scheduling function. The modified architecture proposes to add W[i-15] and W[i-14] before passing to the message scheduling function (sub-module) wordaddm. This brings a hardware reduction of adders in the message scheduler stage as shown in Fig. 23.

Implementation of Improved High Speed SHA-256 Algorithm …

11

Fig. 18 SHA-256 modified message compression block diagram

Fig. 19 Generation of newE in modified message compression block diagram

5 Reducing 11 Adders to 5 Adders in Message Compression In the Fig. 10, the addition of H[ii], W[ii], K [ii] are independent to each other and also to the message compression function. The modified architecture proposes to add W [ii], H[ii] and K [ii] before passing to the message compression() sub-module. This brings a hardware reduction of adders in the message compression stage as shown

12

Fig. 20 Generation of newA in modified message compression block diagram

Fig. 21 Modified generation of final hash (Message Digest)

Fig. 22 SHA-256 modified message scheduler block diagram

B. Srikanth et al.

Implementation of Improved High Speed SHA-256 Algorithm …

13

Fig. 23 SHA-256 modified word generation unit block diagram

in Fig. 24. Also the storage of resultant ((SumE + ch) + h_w_k) in an intermediate variable and using that variable in calculation of newA and newE would reduce the count of adders from 11 to 5.

sumE

ch

+ clk

+

H_W_K

A[ii+1]=newA

A[ii]=a B[ii+1]=newB

B[ii]=b C[ii]=c

C[ii+1]=newC

Intermediate D[ii]

D[ii]=d D[ii+1]=newD E[ii]=e

Message Compresor E[ii+1]=newE

F[ii]=f

F[ii+1]=newF

sumA

+

G[ii]=g H[ii]=h

G[ii+1]=newG

H_W_K

H[ii+1]=newH

+ maj

newE

Note:

. . .

ii ranges from 0 to 63 H_W_K=H[ii]+W[ii]+K[ii] + is carry save adder

+

newA

Fig. 24 SHA-256 modified and updated message compression block diagram

14

B. Srikanth et al.

Fig. 25 Simulation results

Fig. 26 Simulated waveform of message scheduler function

6 Results of Simulation and Synthesis Reports 6.1 Simulation Results In Xilinx, the simulation is performed and the following waveform is observed. Input = “abc” as shown in Figs. 25, 26 and 27. Output = “BA7816BF 8F01CFEA 414140DE 5DAE2223 B00361A3 96177A9C B410FF61 F20015AD”.

6.2 Synthesis Results The synthesis results are generated by using Cadence Genus Synthesis Tool using gpdk 45 nm technology as shown in Table 1, Figs. 28, 29 and 30.

Implementation of Improved High Speed SHA-256 Algorithm …

15

Fig. 27 Simulated waveform of message compressor function Table 1 Comparison between conventional SHA-256 and optimized SHA-256 algorithms Parameter

Conventional SHA-256

Optimized SHA-256

% Improvement

Frequency (Mhz)

181.8

200

10.011

Leakage power (nW)

124,380.906

109,069.468

14.0382

Dynamic power (nW)

2,053,325,483.079

1,539,012,358.983

33.4184

Total power (nW)

2,053,449,863.985

1,539,121,428.451

33.417

Cells

120,440

128,972

–6.61539

Fig. 28 Graphical analysis on frequency

16

B. Srikanth et al.

Fig. 29 Graphical analysis on area

Fig. 30 Graphical analysis on power

7 Conclusion The SHA 256 algorithm generates the 32 byte signature of text file. The optimized SHA-256 implementation achieved 10% improvement in frequency, 14.0382% improvement in leakage power, 33.4184% improvement in dynamic power, 33.417% improvement in total power as compared with conventional SHA-256 even though the area 6% is increased.

References 1. Bai, L., Li, S. (2009). VLSI Implementation of High-speed SHA-256”. IEEE. 2. Mahesh, T. R., Vinoth Kumar, V., Shashikala, H. K., & Roopashree, S. (2023). ML algorithms for providing financial security in banking sectors with the prediction of loan risks. In Artificial Intelligence and Cyber Security in Industry 4.0 (pp. 315–327). Singapore: Springer Nature Singapore. 3. Manasa, K., & Leo Joseph, L.M.I. (2023). IoT security vulnerabilities and defensive measures in industry 4.0. In V. Sarveshwaran, J. IZ. Chen & D. Pelusi (Eds.) Artificial Intelligence and Cyber Security in Industry 4.0. Advanced Technologies and Societal Change. Singapore: Springer. https://doi.org/10.1007/978-981-99-2115-7_4.

Implementation of Improved High Speed SHA-256 Algorithm …

17

4. Lakshmanna, K., Shaik, F., Gunjan, V. K., Singh, N., Kumar, G., & Shafi, R. M. (2022) Perimeter degree technique for the reduction of routing congestion during placement in physical design of VLSI circuits. Complexity, 2022, 11 p. https://doi.org/10.1155/2022/8658770. 5. Kavitha, A., et al. (2022). Security in IoT Mesh Networks Based on Trust Similarity. IEEE Access, 10, 121712–121724. https://doi.org/10.1109/ACCESS.2022.3220678 6. Shaik, F. (2022). An enhanced image processing model for earlier detection and analysis of diabetic foot hyperthermia through cognitive approach. In V K Gunjan, J M Zurada (Eds.), Modern Approaches in Machine Learning & Cognitive Science: A Walkthrough. Studies in Computational Intelligence (Vol. 1027). Cham: Springer. https://doi.org/10.1007/978-3-03096634-8_48. 7. Karimullah, S., & Vishnu Vardhan, D. (2022). Pin density technique for congestion estimation and reduction of optimized design during placement and routing. Applied Nanoscience. 8. Li, C., Wei, Y., & Sun, B. (2008). Structure security of compress function of SHA-256. Journal of Applied Sciences-Electronics and Information Engineering, 26(1), 1–5. 9. Michail, H., & Kakarountas, A. P. (2005). A low-power and high-throughput implementation of the SHA-l hash function. In IEEE International Symposium on Circuits and Systems, ISCAS 2005 (Vol. 4, pp. 4086–4089). 10. Prasad, P. S., Sunitha Devi, B., Janga Reddy, M., & Gunjan, V. K. (2019). A survey of fingerprint recognition systems and their applications. In A. Kumar & S. Mozar (Eds.), ICCCE 2018. Lecture Notes in Electrical Engineering (Vol. 500). Singapore: Springer. https://doi.org/10. 1007/978-981-13-0212-1_53. 11. Huang, Z. (2005) Efficient hardware architecture for secure hash algorithm SHA-l. Journal of Tsinghua University (Science and Technology), 45(1). 12. Karimullah, S., Vishnuvardhan, D., Riyazuddin, K., Prathyusha, K., & Sonia, K. (2021). Low power enhanced leach protocol to extend WSN lifespan. In ICCCE 2020 (pp. 527–535). Singapore: Springer. 13. Gunjan, V. K., Kumar, A., & Rao, A. A. (2014). Present & future paradigms of cyber crime & security majors—growth & rising trends. In 2014 4th International Conference on Artificial Intelligence with Applications in Engineering and Technology, Kota Kinabalu, Malaysia (pp. 89–94). https://doi.org/10.1109/ICAIET.2014.24. 14. Potula, S.R., Selvanambi, R., Karuppiah, M., & Pelusi, D. (2023). Artificial intelligence-based cyber security applications. In: V. Sarveshwaran, J. I. Z. Chen, D. Pelusi (Eds.), Artificial Intelligence and Cyber Security in Industry 4.0. Advanced Technologies and Societal Change. Singapore: Springer. https://doi.org/10.1007/978-981-99-2115-7_16 15. Fahimuddin, S., Subbarayudu, T., Vinay Kumar Reddy, M., Venkata Sudharshan, G., & Sudharshan Reddy, G. (2023). Retinal boundary segmentation in OCT images using active contour model. In A. Kumar, S. Senatore, V.K. Gunjan (Eds.), ICDSMLA 2021. Lecture Notes in Electrical Engineering (Vol. 947). Singapore: Springer. https://doi.org/10.1007/978-981-19-59363_82

Effect of Social Networking Advertisements (SNAs) on Attitudes and Purchase Intention Towards Brand Products Venkata Subbiah Potala and A. S. Sathish

1 Introduction In current scenario, as firms expand their ad campaigns to include mobile phone and internet promotions, there is a necessity for a knowledge on a commercial acquaintance outline that isn’t limited to television advertising as well as other social networking sites. Social networking users pay monetary benefits to display an advertising in their websites and their platforms [1]. In the same way alongside other content on the websites when they visit them. For millions of individuals all around the world. Now a days social networking sites in websites has paly a vital role and its impact on attitude of the consumers are more which intern to impact on decision making of the consumers [2]. The primary social networking sites such as six degree.com as well as myspace, LinkedIn and other popular social networking site Facebook…etc. which are tolerates the users to develop their own profiles and their separate interface with exiting and modern contracts, users can also their views and knowledge sharing through this social networking sites popularity has been increase, know it is considers as a subdivision of social media [3]. Professional website elements like as graphical information, color-coded pertinent data, and audio communication boost website accessibility, and the bulk of these features are found in almost every Networking site. Social networking sites have facilitated consumer-toconsumer communication and enabled customer-brand engagement [4]. As a result of society’s rapid acceptance of social media, social interacting blogs has become an integral component of modern life. Advertisements on social networking sites have V. S. Potala (B) GITAM School of Business, GITAM (Deemed to be University), Visakhapatnam, AP, India e-mail: [email protected] A. S. Sathish VIT Business School, VIT University, Vellore, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 V. K. Gunjan et al. (eds.), Modern Approaches in Machine Learning and Cognitive Science: A Walkthrough, Studies in Computational Intelligence 1117, https://doi.org/10.1007/978-3-031-43009-1_2

19

20

V. S. Potala and A. S. Sathish

given advertisers access to a new network via which they may communicate with their target demographic clients. In social media two types of marketing strategies are more popular such as Facebook and twitter in addition to this YouTube is also one of the popular social networking platforms in order to sharing the videos. In second the paid network media such as advertising through Facebook and advertising through the blogs. Nevertheless, according to Nielsen, social networking sites has renowned and it has free tools has popularity in order to promote the products through the social networking platforms [5].

2 Review of Literature The revolutionary of social networking sites impact has been increases and this networking sites marketing strategies not only gives an exclusive phase to the companies for to promote the company products and services. But also considered this social media platform as a tool for seeing the customer suggestions and feedbacks. A. A. Alalwan is a writer who lives in the United Arab Emirates (2018). According to Maclaran and Catterall, virtual community research is important because of the growing number of participation clusters scheduled the Cyberspace, and discussion forums, which are often market orientated, can provide excellent understandings for vendors [6]. This social networking sites affords sharing the electronic information and also Offred the to maintain long lasting connections. In social networking sites scenario, most of the companies are using social media in the form of online as well as offline, they are developing a unique Facebook group in ordered to promote their products through the social media platforms such as Facebook in accumulation to the banner advertisings [7]. This main purpose of this social networking sites is providing updated information about the products or services and proving the videos with regards initiation of the products and services as well as information with regards the unique promotional deals. The main aim of the social networking sites is to provide information as well as marketing communication which are Offred by the organizations marketing strategies are more popular such as Facebook and twitter in addition to this YouTube is also one of the popular social networking platforms in order to sharing the videos. In second the paid network media such as advertising through Facebook and advertising through the blogs. Social networking sites clients’ users and followers of specific companies are encouraged to take part in frequent conversations on these followers networking sites which are help develop a community where people can share their customer engagement with each other [8]. The company’s SNAs became viral as a result. The major goal of using social networks for advertising reasons is to build brand recognition, as per the report gave by the Nielson in the year of 2013, Venders have underway allocating particular budgets aimed at sponsored social media marketing, according to the report, and want to increase the expenditure in the future years [9]. Marketers may utilize social

Effect of Social Networking Advertisements (SNAs) on Attitudes …

21

media to develop relationships with current and potential customers, as well as create communities where people can cooperate to solve problems (2013). Experts argue that social networking platforms, which are used by millions of people from all over the world from different republics, various languages, and from the various demographical factors, its strength be a good place to promote. Because of marketers utilize both free and paid techniques to promote their products on social media sites, it’s crucial to figure out which features of these ads get the most attention from users [10]. The inclination to respond positively or negatively to a highlight that is rising throughout a certain exposure event is referred to as attitude toward advertisement. Both social networking sites and marketers’ profit from these adverts. The attitude and behaviour as well as behavioral intention of the consumers are having inter connected [11]. In this way this social networking sites will more impact on attitude and it also influence on behavior of the consumers. Recognizing users’ attitudes toward SNAs on social networking sites is critical for both marketers and SNS managers. In order to determining the effectiveness and impact of attitude on behavioral intention of the various advertisings in social media and advertising studies. As per current scenario these two are plays a major goal in advertising research. Promotional messages may spread via electronic word-of-mouth if people have a positive attitude toward social networking sites (e-WOM). The social networking sites as well as blogs are are using by the clients in various category in the form of sharing the positive views with respect to products and services [12]. Ballantine has found that well-adjusted blog assessments are professed as the most trustworthy, and credible reviews resulted in the top scores on brand attitude and buy intention. Individuals’ interactions and connections with one another, sharing similar interests, participating in discussion forums, and sharing their opinions have all changed as a result of social networking sites. Consumers’ attitudes and behaviour toward products and services are influenced through word of mouth, which is the exchange of advertising messages among customers. According to Lorenz, the television can influence client purchase decisions at any point in the process. According to studies, social media networks like Facebook can help determine client purchase intent. As a result, brand advertisers must design and implement a flawless strategy while building social networking sites in order to elicit buy intent from social networking manipulators engaging social networking-based advertisings for any branded products [13, 14]. To reach their target market, brand managers employ a range of mediums. Social networking sites afford to the vendors with a reliable stage for appealing with social networking sites operators and communicating the brand’s fundamental message [15]. Promotion should increase client perceptions of the organisation and its products, inspire behavioral responses such as positive word-of-mouth, and elicit a desire to buy. M. Ali and S. A. Raza, to elicit such psychological—behavioral responses, advertising should offer information that is relevant or appropriate to the promoted brand, fun and compelling to watch or read, and reliable enough for people to trust

22

V. S. Potala and A. S. Sathish

[16]. The current study has demonstrated that the impact of social networking advertisements varies based on the stage of a user’s purchase decision making, when they are buying a product or service from various social media networks [17, 18].

3 Objectives of the Study • To identify the influence of social networking sites (SNS) on consumer attitude. • To measure the influence of social media or social networking sites on consumer at different stages of purchase. • To know Demographical Impact on purchase Intention of the consumers.

4 Significancy of the Study Considering the purposes of the research, performing the investigation from the consumer’s perspective would be the best strategy. This study is being carried out by the researcher in order to assist customers in determining the reasons why social media influenced their purchasing decisions. Because the primary goal of marketing is to analyses the demands of the consumer, the data acquired through the questionnaire is from the consumer’s perspective in order to identify fresh insights. The study also intends to assist potential readers comprehend the role of social media websites/ apps in consumer decision-making. The study is majorly emphasized on end-user behaviour.

5 Research Methodology In research methodology, we not only discuss methodological approaches, but also the logic behind methodologies we use in the content of our research study and explain why we are using a particular method or technique; thus, in this investigation, we will look at various steps that are commonly used in studying and the logic behind them. It’s a high-level overview of the study’s methodology and procedures. The study used a descriptive research approach. Surveys and different forms of fact-findings were examples of descriptive data. A statistical measure for determining the correlation or co-relationship between two variables is correlation. Regression is the study of how to link an independent variable to a dependent variable statistically.

Effect of Social Networking Advertisements (SNAs) on Attitudes …

23

A self-administered questionnaire survey based on the findings of the literature review was created to achieve the study’s purpose. Following the pre-test, the questionnaire was improved. The survey was organized into four sections, each dealing with a distinct topic: Tsiros et al. identified two types of media: (1) social media, and (2) traditional media (advertising).

6 Conceptual Model See (Fig. 1).

7 Data Analysis For social sciences, a statistical software is available (SPSS) For a stepwise regression study, statistics has generated quite a few tables of output. In this part, the researcher simply considers the three primary tables needed to evaluate the stepwise regression procedure’s results, provided as shown in Fig. 1 that no assumptions have been violated. When reviewing data for the two assumptions necessary to do stepwise regression analysis, a thorough explanation of the outcome must be understood. Relevant scatterplots, histograms (with overlay normal curves), Normal P-P Plots, case-by-case diagnostics, and the Durbin-Watson statistic are all included. Only the findings of the stepwise regression analysis are discussed here. The impact of social networking sites (SNS) on predictors to consumer attitude R and R square values seen in Table 1. The R value denotes simple correlation and is 0.790, indicating a high degree of correlation. The R square value which represents that number reflects how much the independent variable, Entertainment, Information, can explain in terms of the overall variance in the dependent variable, Consumer Attitude. 79% of the variance can be accounted in this situation, which is a significant as shown in Table 2.

Fig. 1 Conceptual model

24

V. S. Potala and A. S. Sathish

Table 1 Model summary step wise regression Model

1

R

0.790a

R2

0.617

Adjusted R2

Std. error

0.617

0.99369

Change statistics R square change

F change

df1

df2

Sig. F change

0.617

292.193

1

188

0.000

a Predictors: (Constant), Attitude

Table 2 ANOVA a step wise regression with significant values Model 1

2

Sum of squares

df

Mean square

F

Sig.

Regression

138.175

1

138.175

87.008

0.000b

Residual

298.556

188

1.588

Total

436.730

189

Regression

156.891

2

78.445

52.420

0.000c

Residual

279.840

187

1.496

Total

436.730

189

The regression analysis positively predicted the dependent variable, as shown in Tables 3, 4, 5, 6, 7, and 8. This table extracted worth signifies that the regression model’s got statistical significance values. The significant value p value is less than 0.05, indicating that the regression model statistically substantially predicts the outcome variable. The impact of social networking sites SNS on consumer attitude to word of mouth See (Tables 4, 5 and 6). The impact of social networking sites SNS on consumer attitude to purchase intention Stepwise regression, as shown in Table 9, is a mix of forward and backward selection procedures. It used to be highly popular, but the Multivariate Variable Selection process described later in this chapter will always perform at least as well, if not better. The Coefficients table above provides us with the knowledge we need to anticipate pricing based on income and determine if income has a statistically significant influence on the models (Tables 7, 8 and 9).

3.488

0.391

0.283

(Constant)

Entertainment

Information

4.301

0.507

(Constant)

0.517

0.080

0.062

0.552

0.054

0.244

0.433

0.562

Beta

B

Std. error

Standardized coefficients

Unstandardized coefficients

Entertainment

a D.V: Attitude

2

1

Model

Table 3 Coefficientsa step wise regression

3.536

6.276

6.316

9.328

8.312

t

0.001

0.000

0.000

0.000

0.000

Sig.

0.474

0.562

0.562

Zero-order

Correlations

0.250

0.417

0.562

Partial

0.207

0.367

0.562

Part

0.719

0.719

1.000

Tolerance

1.390

1.390

1.000

VIF

Collinearity statistics

Effect of Social Networking Advertisements (SNAs) on Attitudes … 25

26

V. S. Potala and A. S. Sathish

Table 4 Model summary step wise regression Model

1

R

0.780a

R square

0.617

Adjusted R square

Std. error

0.606

0.99369

Change statistics R square change

F change

df1

df2

Sig. F change

0.617

292.193

1

188

0.000

a P: (Constant), Attitude

Table 5 ANOVAa step wise regression Model 1

Regression

Sum of squares

df

Mean square

F

Sig.

288.517

1

288.517

292.193

0.000b

0.987

Residual

185.635

188

Total

474.152

189

a D.V: WOM b PR (Constant), Attitude

8 Findings of the Study The study’s intention was to identify the how social media advertising affected people’s opinions about social networking ads, as well as the influence of word of mouth and purchase intent after seeing social media ads in Karachi. A total of 500 Indian social media users’ questionnaires were found to be beneficial. Stepwise regression analysis was used to examine the associations. We also noticed that has a positive impact on the value of social media advertising. The majority of responses were between the ages of 15 and 25, as this is the age group that spends the most time on social media. According to poll results, about half of all consumers purchase online once or twice per month. This might imply that they are from the working and middle classes, who do not have much spare time and who tend to buy in large quantities but less frequently from online platforms. According to the poll results, about half of the respondents spend 0–2 h every day on Social Media Networks.

1.819

0.813

Attitude

0.048

0.436 0.780

Beta

B

S. E

S. coefficients

Unstandardized coefficients

(Constant)

a D.V: Word of mouth

1

Model

Table 6 Coefficientsa step wise regression

17.094

4.169

t

0.000

0.000

Sig.

0.780

Zero-order

Correlations

0.780

Partial

0.780

Part

1.000

Tolerance

1.000

VIF

Collinearity statistics

Effect of Social Networking Advertisements (SNAs) on Attitudes … 27

28

V. S. Potala and A. S. Sathish

Table 7 Model summary-step wise regression Model

1

R

0.731a

R square

0.535

Adjusted R square

Std. error

0.532

1.05046

Change statistics R square change

F change

df1

df2

Sig. F change

0.535

216.105

1

188

0.000

a P: (Constant), Attitude

Table 8 ANOVAa step wise Model 1

Regression

Sum of squares

df

Mean square

F

Sig.

238.465

1

238.465

216.105

0.000b

1.103

Residual

207.451

188

Total

445.916

189

a D.V: Purchase Intention b P: (Constant), Attitude

Customers follows brands on social media because they want to hear about deals, new items, and so on. According to the survey results, about 61 percent of individuals use social media as an electronic word of mouth, and many consumers base their purchasing decisions on social media referrals. According to the study, over 47 percent of respondents feel that the brand’s social media page/website influenced their perception of it. Most internet shoppers read evaluations about the brands of the items they are considering purchasing, which may influence their perception of that brand. According to the survey results, over 88 percent of respondents say social media plays an essential role in brand promotion since it increases visibility, allowing the company to create leads and improve sales.

2.491

0.739

(Constant)

Attitude

0.461

0.050

0.731

Beta

B

Std. error

Standardized coefficients

Unstandardized coefficients

a D.V: Purchase Intention • D.V: Mean that Depending variable • P: Mean that Predictor

1

Model

Table 9 Coefficientsa step wise regression

14.701

5.400

t

0.000

0.000

Sig.

0.731

Zero-order

Correlations

0.731

Partial

0.731

Part

1.000

Tolerance

1.000

VIF

Collinearity statistics

Effect of Social Networking Advertisements (SNAs) on Attitudes … 29

30

V. S. Potala and A. S. Sathish

9 Conclusion The conclusion of this study customers in the study area are vigorously by means of social television stages to verify their purchasing decisions. The mainstream of defendants contemplates that social media to be an arrangement of word-of-mouth will give more impact on consumer attitude. Previous customers’ recommendations and preferences stated on social media platforms impact new customers’ decisionmaking. When compared to those who relied on other sources of information, television users found decision-making to be simpler and more fun. Those who felt social media content was of better quality and quantity than expected were generally pleased. Inclusive, the data show that social media has a substantial impact on user behaviour.

10 Future Direction of the Study Contempt to the fact that the study remained conducted following strict protocols, it includes a number of flaws. Social media marketing is not a new trend, and it continues to develop and evolve. There are a number of articles and books on the issue, but few of them link social media to changes in consumer behaviour. The results of the study are described in broad strokes. The study’s data sample was exceedingly small, implying that its generalizability was restricted. To safeguard more individuals in society and achieve more reliable results, the sample size should be increased. Due to differences in culture and values, consumer purchase behaviour may vary from country to country. The objectives have been narrowed to increase the study’s concentration, and the research only provides information on what has to be researched in light of the goals. Future research might use different scales and compare the findings to those found in this study. The application of this model in future experimental investigations is supposed to underwrite to the existing literature with respect to social networking sites impacts on users. In order to improve the subject of the study, it should be applied to a variety of client groups with different socio-demographic characteristics.

References 1. Alalwan, A. N. (2018). Investigating the impact of social media advertising features on customer purchase intention. International Journal of Information Management, 42, 65–77. https://doi. org/10.1016/j.ijinfomgt.2018.06.001. 2. Shaik, F., Giri Prasad, M. N, Rao, J., Abdul Rahim, B. & Soma Sekhar, A. (2010). Medical image analysis of electron micrographs in diabetic patients using contrast enhancement. In 2010

Effect of Social Networking Advertisements (SNAs) on Attitudes …

3.

4.

5.

6.

7.

8.

9.

10. 11. 12.

13.

14.

15.

31

International Conference on Mechanical and Electrical Technology, Singapore (pp. 482–485). https://doi.org/10.1109/ICMET.2010.5598408. Rashid, E., Ansari, M. D., Gunjan, V. K., & Ahmed, M. (2020). Improvement in extended object tracking with the vision-based algorithm. In V. Gunjan, J. Zurada, B. Raman & G. Gangadharan (Eds.), Modern Approaches in Machine Learning and Cognitive Science: A Walkthrough. Studies in Computational Intelligence (Vol. 885). Cham: Springer. https://doi.org/10.1007/9783-030-38445-6_18. Palsodkar, P., Palsodkar, P., Gokhale, A., Dorge, P., & Gurjar, A. (2022). Fuel Larceny and Leakage Indication System Using IoT. In Intelligent Systems for Social Good: Theory and Practice (pp. 81–89). Singapore: Springer Nature Singapore. Ballantine, P., & Y. C. (2015). The effects of review valence in organic versus sponsored blog sites on perceived credibility, brand attitude, and behavioural intentions. Marketing Intelligence & Planning, 33, 508-521. https://doi.org/10.1108/MIP-03-2014-0044. Shaik, F., Sharma, A. K., & Ahmed, S. M. (2016). Detection and analysis of Diabetic Myonecrosis using an improved hybrid image processing model, in 2016 2nd International Conference on Advances in Electrical, Electronics, Information, Communication and BioInformatics (AEEICB), Chennai, India (pp. 314–317). https://doi.org/10.1109/AEEICB.2016. 7538298. Bart, Y., Stephen, A., & Sarvary, P. (2014). Which products are best suited to mobile advertising? A field study of mobile display advertising effects on consumer attitudes and intentions. Journal of Marketing Research., 51, 270–285. https://doi.org/10.1509/jmr.13.0503 Rudra Kumar, M., Pathak, R., & Gunjan, V. K. (2022). Machine learning-based project resource allocation fitment analysis system (ML-PRAFS). In A. Kumar, J. M. Zurada, V. K. Gunjan & R. Balasubramanian (Eds.), Computational Intelligence in Machine Learning. Lecture Notes in Electrical Engineering (Vol. 834). Singapore: Springer. https://doi.org/10.1007/978-981-168484-5_1. Senthil Kumar, G., Ramana, K., Mohana, R. M., Aluvalu, R., Gunjan, V. K., & Singh, N. (2022). An effective bootstrapping framework for web services discovery using trigram approach. Mobile Information Systems, 2022, Article ID 2373596, 12 p. https://doi.org/10.1155/2022/ 2373596. Gallaugher, J., & Ramsbotham, S. (2010). Social media and customer dialog management at starbucks. MIS Quarterly Executive, 9. Gupta, M. (2020). A study on consumer buying behaviour for shopping FMCG in large retail outlet. ANVESHAK-International Journal of Management, 7(2), 82–94. Castano, E., Yzerbyt, V., Paladino, M. P., & Sacchi, S. (2002). I belong, therefore, I exist: Ingroup identification, ingroup entitativity, and ingroup bias. Personality and Social Psychology Bulletin, 28(2), 135–143. Kalbande, M., Gaidhani, Y., Panse, T., & Mahajan, M. (2022). Cloud based examination hall authentication system using fingerprint module. In S. Mukherjee, N. B. Muppalaneni, S. Bhattacharya & A. K. Pradhan (Eds.). Intelligent Systems for Social Good. Advanced Technologies and Societal Change. Singapore: Springer. https://doi.org/10.1007/978-981-19-0770-8_15. Gunjan, V. K., Prasad, P. S., Fahimuddin, S., & Bigul, S. D. (2019). Experimental investigation to analyze cognitive impairment in diabetes mellitus. In: A. Kumar & S. Mozar (Eds.), ICCCE 2018. Lecture Notes in Electrical Engineering (Vol. 500). Singapore: Springer. https://doi.org/ 10.1007/978-981-13-0212-1_79. Gunjan, V. K., Kumar, A., & Rao, A. A. (2014). Present & future paradigms of cyber crime & security Majors—Growth & Rising trends. In 2014 4th International Conference on Artificial Intelligence with Applications in Engineering and Technology, Kota Kinabalu, Malaysia (pp. 89–94). https://doi.org/10.1109/ICAIET.2014.24.

32

V. S. Potala and A. S. Sathish

16. Ferrara, E., Varol, O., Davis, C., Menczer, F., & Flammini, A. (2016). The rise of social bots. Communications of the ACM, 59(7), 96–104. 17. Fantin Irudaya Raj, E., & Appadurai, M. (2022). Internet of things-based smart transportation system for smart cities. In Intelligent Systems for Social Good: Theory and Practice (pp. 39–50). Singapore: Springer Nature Singapore. 18. Raja, P. (2022). ML in WSN Using IoT for smart cities: A survey. In Intelligent Systems for Social Good: Theory and Practice (pp. 1–11). Singapore: Springer Nature Singapore.

Design of Optimal Waste Management System Using IOT and Machine Learning Technique in Educational Institutions L. Sivayamini, C. Venkatesh, and Fahimuddin Shaik

1 Introduction In Modern technologies IoT is mostly used technology to invent the new technologies for the use of society. Now we are going to do a project on the waste management system by using IOT and Machine learning techniques. Because of its broad connection, the Internet of Things (IoT) is a new and intriguing technology that has the potential to positively impact human life [1]. The Internet of Things (IoT) allows low-energy devices to communicate and interact with each other. Based on the background of IoT, many applications all over the world have implemented various processes. To prevent the wastage in the society this technology we are proposing. In the streets, near schools, colleges, outside the cities wherever we go we are seeing wastage anywhere [2]. Many people are working in the society for the collection of waste. People they are going to collect the waste from the houses, near drainages, and in the markets. They will take that waste into the vehicles and they will throw out of the villages. Some people those who are not working property in the municipality the head officers doesn’t know about the people working properly or not [3]. To maintain the proper collection of waste and to find the proper working of the workers this is very easy and useful to find the collection of waste. And in this the data will be seen only by the head officers in the corporation.

L. Sivayamini · C. Venkatesh (B) · F. Shaik Department of ECE, AITS, Rajampet, AP, India e-mail: [email protected] L. Sivayamini e-mail: [email protected] F. Shaik e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 V. K. Gunjan et al. (eds.), Modern Approaches in Machine Learning and Cognitive Science: A Walkthrough, Studies in Computational Intelligence 1117, https://doi.org/10.1007/978-3-031-43009-1_3

33

34

L. Sivayamini et al.

Most of the times the waste is everywhere in the society. And there are so many dump yards are available but people are not using properly, So to maintain the neatness, clean in the society the municipality workers are working for the sake of clean society [4]. Now a days the workers also not working properly they are doing their own works at the time of office hours. And they are collecting waste in some areas and dumping that waste into the dump yards and they are moving to their own works, So to identify the work of the workers and the collection of waste how much quantity they are dumping daily to measure these this waste management system we are proposing [5]. Even though previously so many techniques are used to find the quantity of waste and working of the people based on the machine learning and GSM techniques. In those the data will not be secure. To secure the data of the collection of wastage this system we are proposing [6].

2 Literature Survey Many academics have researched and developed innovative applications for smart cities, particularly in the area of waste management based on the benefits of IoT technologies. A simple system was demonstrated that detects the fullness of garbage cans. Which collected data and sent it using a wireless mesh network to save electricity and increase operational time. However, the system still has some confusing issues with the concept [7]. Software platform for smart cities was proposed to improve waste management; however, they primarily focused on Mobile Computing data collecting, and their platforms were made up of technologies from other firms. Some approaches, on the other side, optimization-based waste management solutions have been established [8]. In order to achieve a more efficient system. Using Lora WAN technology and route optimization, the authors suggested stage of trash management and control that might be used in some locations. In, a technique for collecting food waste was developed, in which data was collected using radio frequency identification (RFID) technology and relayed via a mesh network of wireless devices. However, this technology has long-term limitations. Were significant, especially given the smart city’s goal of large-area control. Finally, the optimization algorithm’s results grew too to be applied to a real-world system like a city, it’s a little fuzzy [9]. In, an outstanding architecture for a sensor node was suggested, which combined a microcontroller (ATMega328P) and a TTL—100 433 MHz module for practical waste management. Nonetheless, they simply used this platform is used to offer sensor nodes, and they did not use waste management techniques in the process. While the goal the purpose of this essay was to propose an Internet of Things application using the microcontroller board design was quite hard also, certain performance and functionality are required for each application. Our research uses machine learning and graph theory to improve waste collection systems, eliminate overflowing bins, and cut labour costs. In this study, we look at algorithms based on heuristic models or graph theory that can help us uncover strategies to reduce garbage collection distance [10]. The main goal is to minimise total transportation costs by transferring, saving labour, and

Design of Optimal Waste Management System Using IOT and Machine …

35

reducing reliance on obsolete cars, all while improving service quality and general quality of life. The other programme, which employs machine learning and graph theory techniques, prioritises garbage collection over cost and energy savings, and aids the university’s waste management network in performing at its best [11]. We compare and contrast the properties of the algorithms discussed in this part with our method. Our technology connects the Smart Dust Bins to the internet so that we can get real-time data on them [12, 13]. These bins are linked to an ultrasonic sensor-equipped Raspberry Pi system. The ultrasonic sensor measures the amount of dust in the dustbin and transmits signals to the Raspberry Pi, which encodes and delivers the signal to the application, which receives it [14]. The data was received, examined, and processed in a database that shows the state of garbage in the trashcan based on an authorized person’s request. The appropriate authority is notified that the dustbin is full, and the person in charge of collecting waste from the specific locations [15, 16].

3 Existing Methodology The current approach can only measure the food waste of an entire organization or a specific bin for a single day, not for each and every individual person. With that information, the organization can only take a few steps to prevent food waste. The current approach is a semi-automated procedure in which the report must be personally checked and the incentives must be manually selected. As previously stated, the current technology will only calculate the food waste of a whole organization. As a result, the organization’s data is only used in a limited number of ways. Even with the obtained data, there would be little influence in raising awareness or taking action among the organization’s employees. The current system is deficient in the data collecting area, and the applications that can be created with the data collected are limited. As a result, the current system has limited utility. This method uses a mobile application to monitor garbage bins and provide information on the amount of rubbish collected. The system detects the rubbish level and compares it to the depth of the garbage containers using ultrasonic sensors positioned over the bins. The Raspberry Pi is used to deliver data in this system. A 12 V power supply is used to power the system. The purpose of an application is to show the status to the person who is monitoring it. The programme shows a view of the garbage bins and color-codes the waste collected to display the amount of garbage collected. The data is saved in the database that is produced, and then it is retrieved. Communication is critical for service provisioning in IoT applied to external and public contexts. Because this sort of GSM has such a broad service domain, devices must be able to communicate with one another in a secure manner. As a result, the proposed system’s SGBs communicate with one another via a wireless mesh network, ensuring communication dependability. These equipment may need to move in the external environment on occasion. The suggested system’s mobility is ensured via a battery-based power supply.

36

L. Sivayamini et al.

4 Proposed Methodology The main aim of the proposed methodology is to collect the waste based on the quantity through the RFID cards provided to the each and every vehicle. And that will displays the time at what time they are dumping the waste into the dump yards. For the different type of vehicles they will keep workers by using with this system the workers also do their work properly within the given hours. For this we are proposing the new methodology called IoT and machine learning. By using the IoT the data will be secure and that data can be only read by the higher authorities in the corporations. We will use the machine learning code by using embedded c for the Arduino and then it is converted into the machine learning by using raspberry pi. And the proposed method will have the so many advantages it will reduces the cost. The data will be very secure. For this data secure we use the server called Think Speak in this the data will give for the different vehicles. By using these different cards for each vehicle will be entered into the server and the quantity that the vehicle is dumping the waste into the dump yards. In the below block diagram shows the circuit connections made on the board. Firstly we are using the two controllers those are raspberry pi3 and Arduino these are the major working mechanism for the circuit. ADC is used to convert the analog input to the digital input that is the convertor. We kept the load in the form of analog input but we need the results in the digital form so we need to use the ADC. The load cell is use to keep the load on the cell here we are proposing the weight limit of the circuit upto 3kgs. To dump the program into the controller the power supply is used by using this power supply to the raspberry pi3 the inbuilt code will be generates in our system. Now we will keep the load on the load cell and here the RFID reader is used to scan the vehicle number. The government will allocate different numbers or ids for each vehicles. If one vehicles arrives that vehicle will load the waste and by using RFID reader it will scans the card and it will takes the weight and displays how much load they were added. Cloud /server is used to enter the whole information of each and every vehicle into the server and that was only seen by the head of the corporation. Think Speak server is connected through our cloud/server then our data is stored into the server it will displays id of the vehicle, weight, timings, date as shown in Fig. 1. If we add the weight of the same vehicle again by using RFID card it will add the data to the previous amount of the weight. If we didn’t gave any load means it will shows invalid. The above Fig. 2 shows the block diagram for the power supply. Step down transformer is used to maintain the voltage powers we need the lower voltages for the kit it will converts the higher voltage powers to the lower voltage powers. Our system requires low voltages. And as the inputs we are providing ac inputs to convert into dc the bridge rectifier is used. Regulator is provided to keep the constant voltages. Then finally that will provides the output towards the raspberry pi.

Design of Optimal Waste Management System Using IOT and Machine …

37

Fig. 1 Block Diagram for waste management system using IoT and Machine learning

Step down Transf ormer

Brid ge Rect ifier

Filt er

Regu lator

Out put

Fig. 2 Block diagram of Power supply

5 Experimental Results In our proposed model as shown in Fig. 3 the results shows that the managing of the waste based on the usage of different controllers like arduino and raspberry pi3. In the previous chapters we described about the components of the kit. After providing the connections to the raspberry pi and arduino the code is written based on the ids given to the RFID cards. In the above trainer board all components are fixed. By using the load cell the load is weighted on that max load of 3kgs we used. For the different load we used different RFID cards for the particular vehicle or dustbin. The cards will be read by using RFID module. The interconnection between arduino and Raspberry pi by using TTL (Transistor transistor logic). In the above Fig. 3 we kept the load as mobile phone on the load cell. By using RFID module, the card will be scanned. We placed input as analog form that will convert into digital by using ADC convertor. Now the weight will be displayed in the form of digital in grams. The above shown Fig. 4 shows the waste management system using IoT and machine learning. Shows the data that is acquired by the server

38

L. Sivayamini et al.

Fig. 3 Waste management system using IoT and machine learning

Fig. 4 Keeping the load on the load cell as mobile

using RFID reader. It shows the date, timing, entry id and in the fields the weight is displayed. Depicts the graph of Optimal waste management system without any load in it displays the constant and the depicts the graph of Optimal waste management system with load (Waste) it varies accordance with the weight.

6 Conclusion As per the proposed system firstly we will design the circuit as per the block diagram. The load cell is used to weigh the load firstly we will weigh the waste on the load cell and then by using RFID module the card will be scanned the data will be entered into the server by using IoT. In the think speak that will displays the vehicle ID and the quantity of waste that vehicle will be going to dump and time. The data will be

Design of Optimal Waste Management System Using IOT and Machine …

39

very secure that can be read only by the higher authorities. If the weight is added by the same vehicle by using the RFID reader that will displays by adding previous weight to the present weight. By using this proposed method the labours can be easily identified by the head of the corporations whether they are working or not make their work easily by looking everything in the server.

References 1. Silva, B. N., Khan, M., & Han, K. (2018). Towards sustainable smart cities: A review of trends, architectures, components, and open challenges in smart cities. Sustainable Cities and Society, 38, 697–713. 2. Shaik, F., Giri Prasad, M. N., Rao, J., Abdul Rahim, B., & Soma Sekhar, A. (2010). Medical image analysis of electron micrographs in diabetic patients using contrast enhancement. In 2010 International Conference on Mechanical and Electrical Technology, Singapore (pp. 482–485). https://doi.org/10.1109/ICMET.2010.5598408. 3. Davanam, G., Kallam, S., Singh, N., Gunjan, V. K., Roy, S., Rahebi, J., Farzamnia, A., & Saad, I. (2022). Multi-Controller model for improving the performance of IoT networks. Energies, 15, 8738. https://doi.org/10.3390/en15228738 4. Vinagre, E., De Paz, F., Pinto, T., Vale, Z., Corchado, M., & Garcia, O. (2016). Intelligent energy forecasting based on the correlation between solar radiation and consumption patterns. In Proceedings of the 2016 IEEE Symposium Series on Computational Intelligence (SSCI). Athens, Greece: IEEE, December 2016. 5. Siddiquee, K. N.-e-A., Islam, M. S., Singh, N., Gunjan, V. K., Yong, W. H., Huda, M. N., & Bhupal Naik, D. S. (2022). Development of algorithms for an IoT-Based smart agriculture monitoring system. Wireless Communications and Mobile Computing, 2022, Article ID 7372053, 16 p. https://doi.org/10.1155/2022/7372053. 6. Gunjan, V. K., Shaik, F., & Kashyap, A. (2021). Detection and analysis of pulmonary TB using bounding box and K-means algorithm. In A. Kumar & S. Mozar (Eds.), ICCCE 2020. Lecture Notes in Electrical Engineering (Vol. 698). Singapore: Springer. https://doi.org/10.1007/978981-15-7961-5_142. 7. Sukjaimuk, R., Nguyen, Q. N., & Sato, T. (2018). A smart congestion control mechanism for the green IoT sensor-enabled information-centric networking. Sensors, 18(9), 2889. 8. Yellamma, Pachipala, P. G. Sandeep, R. Revanth Sai, S. Rohith Reddy, and D. Mahesh. “Automatic Vehicle Alert and Accident Detection System Based on Cloud Using IoT.“ In Embracing Machines and Humanity Through Cognitive Computing and IoT, pp. 77–85. Singapore: Springer Nature Singapore, 2023. 9. Ahmed, M., Ansari, M. D., Singh, N., Gunjan, V. K., BV, S. K., & Khan, M. (2022). Rating-based recommender system based on textual reviews using iot smart devices. Mobile Information Systems, 2022. 10. Pardhasaradhi, P., Pavan Kumar, K. V. K. V. L., Yathiraju, R., Sumanth, Y. S. S., Nishith, S., & Reddy, T. V. V. (2023). Stuck-At Fault Detection in Ripple Carry Adders with FPGA. In Embracing Machines and Humanity Through Cognitive Computing and IoT (pp. 23–31). Singapore: Springer Nature Singapore. 11. Bharadwaj, A. S., Rego, R., & Chowdhury, A. (2016). IoT based solid waste management system: A conceptual approach with an architectural solution as a smart city application. In 2016 IEEE Annual India Conference (INDICON), Bangalore, India (pp. 1–6). https://doi.org/ 10.1109/INDICON.2016.7839147. 12. Karimullah, S., Vishnu Vardhan, D., & Basha, S. J. (2020). Floorplanning for placement of modules in VLSI physical design using harmony search technique. In ICDSMLA 2019, Lecture Notes in Electrical Engineering (Vol. 601). Springer Nature Singapore Pte Ltd.

40

L. Sivayamini et al.

13. Kavitha, A., et al. (2022). Security in IoT mesh networks based on trust similarity. IEEE Access, 10, 121712–121724. https://doi.org/10.1109/ACCESS.2022.3220678 14. Sai Surya Teja, T., Venkata Hari Prasad, G., Meghana, I., & Manikanta, T. (2023). Publishing temperature and humidity sensor data to ThingSpeak. In M. Usman & X. Z. Gao (Eds.), Embracing Machines and Humanity Through Cognitive Computing and IoT. Advanced Technologies and Societal Change. Singapore: Springer. https://doi.org/10.1007/978-981-19-452 2-9_1. 15. Saha, H. N., Auddy, S., Pal, S., Kumar, S., Pandey, S., Singh, R., Singh, S. K., Banerjee, S., Ghosh, D., & Saha, S. (2017). Waste management using the Internet of Things (IoT). In Proceedings of the 2017 8th Annual Industrial Automation and Electromechanical Engineering Conference (IEMECON), Bangkok, Thailand, August 16–18, 2017 (pp. 359–363). 16. Siddaiah, N., Pardhasaradhi, P., Phanigopi, M., Vasanthi, Y., & Deepika, Y. (2023). Assembly Line Implementation for IOT Applications. In M. Usman & X. Z. Gao (Eds.), Embracing Machines and Humanity Through Cognitive Computing and IoT. Advanced Technologies and Societal Change. Singapore: Springer. https://doi.org/10.1007/978-981-19-4522-9_12.

Detection of COVID-19 Based on Deep Learning Methods: A Critical Review J. Chinna Babu, Bolla Madhusudhana Reddy, K. Swapna, P. Yamuna, K. Sandeep Kumar Reddy, and N. Sumanth

1 Introduction In December 2019, an unidentified animal in Wuhan, China, transmitted the novel coronavirus (COVID-19), which had its origins in bats [1]. COVID-19 was deemed a Public Health Emergency of International Concern by the World Health Organization on January 30, 2020, and a global pandemic on March 11, 2020. Chest X-ray imaging is helpful for a precise diagnosis because the virus initially affects the patient’s lungs [2]. Rapid detection and lowering exposure to the virus among medical or healthcare professionals would both benefit from any automatic, dependable, and accurate screening method for COVID-19 infection [3]. Chest X-rays (CXR) and CT scan images are two examples of radiographic images that have seen extensive use of the deep learning paradigm [4]. These radiographic images contain a wealth of data, including patterns and cluster-like structures that are visible in the conformance and detection of pandemics similar to COVID-19 [5]. The most effective machine learning method is deep learning, which offers insightful analysis for examining a significant number of chest x-ray images that have a significant bearing on Covid-19 screening [6]. The chest X-ray images can be used to classify Covid positive and Covid negative patients using deep learning models. It’s crucial to make a quick diagnosis of COVID-19 and identify high-risk patients with poorer prognoses for early prevention and resource optimization in the medical field. In order to quickly screen for COVID-19 and identify potential high-risk patients, deep learning offers a practical tool that may be useful for the J. C. Babu (B) · K. Swapna · P. Yamuna · K. S. K. Reddy · N. Sumanth Department of Electronics and Communication Engineering, Annamacharya Institute of Technology and Sciences, Rajampet 516126, India e-mail: [email protected] B. M. Reddy Department of EEE, Malla Reddy Engineering College for Women, Hyderabad, India © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 V. K. Gunjan et al. (eds.), Modern Approaches in Machine Learning and Cognitive Science: A Walkthrough, Studies in Computational Intelligence 1117, https://doi.org/10.1007/978-3-031-43009-1_4

41

42

J. C. Babu et al.

efficient use of medical resources and early detection of problems before patients experience severe symptoms [7]. This study investigates how Deep Learning has combated the COVID-19 pandemic and offers recommendations for future COVID19 research. The survey’s main topic is the development of technological solutions, with a particular focus on Deep Learning [8].

2 Literature Review The authors of [1], proposed a concept of Deep Learning in the Detection and Diagnosis of COVID-19 Using Radiology Modalities [9]. The discrepancies across the investigations carried out by the researchers here were addressed by discussion. The study’s name, nation, year of publication, population studied, and study method, and data utilized were among the retrieved data items as shown in Fig. 1. Techniques for DL, methods of assessment, and outcomes. Results After reviewing 160 abstracts and full-text papers, 37 research works that met the inclusion criteria were chosen. In order to pick the articles, the PRISMA approach was

Fig. 1 Block diagram of COVID-19 detection Using Radiology Modalities

Detection of COVID-19 Based on Deep Learning Methods: A Critical …

43

used. All of the chosen publications were published in 2020 because of the disease’s novelty. India had eight articles published, China had five, the United States had five, and Turkey had three, out of a total of 37 articles retrieved. Furthermore, Iran, Italy and Greece submitted two analysis papers on this issue, whereas remaining countries one as shown in Figs. 2, 3, and 4. COVID-19 has been detected, diagnosed, classified, predicted, and prognosed using several DL approaches, according to studies. Domiciliary datasets (counting CT and X-ray scans) as choice popular datasets were utilized in several experiments to train and validate the approaches. Reactivity, precision, and exactness are three criteria for evaluating efficacy of methodologies employed in community research. Nonetheless, AUC has been utilized in a lot of studies to estimate the efficacy about COVID-19 diagnosis approach [10].

Fig. 2 DL based diagnosis and detection modalities of COVID-19

Fig. 3 Radiological modalities of DL based CT and X-Ray Images to detect COVID-19

44

J. C. Babu et al.

Fig. 4 COVID-19 radiological modality pictures were analyzed with a high rate of CNN architectures

Limitations 1. Soft tissue contrast is poor. 2. On a radiograph, delineating soft tissue structures might be challenging or straightforward. The authors of [11], proposed a concept of Deep Learning based detection and analysis ofCOVID-19 ON Chest X-ray images. The details from the Kaggle repository were cleaned to meet the requirements. In order to get trustworthy results from a technique based on deep learning, you’ll need a lot of data. However, it’s probable that every problem, particularly in medical-related ones, lacks sufficient evidence. It can be time-consuming and costly to obtain medical data. Augmentation is a technique that may be used to overcome problems like these. Overfitting can be eliminated by augmentation, and precision can be improved as shown in Figs. 5, and 6. Results Xception Net Typical X-rays of the chest were compared to COVID-19 afflicted persons as part of the results analysis. The accuracy matrices are used to examine Inception Net V3, Xception Net, and resNeXt. After that, the results were compared to see which model was the best. Despite the model’s excellent accuracies, we recommend confirming the results with future dataset updates. Due to a shortage of details, the replica is only instructed on 1560 illustratives. It is an offshoot of the Inception network. Here, depth-wise separable convolutions replace inception modules. It has a framework proportion that is equivalent to inception net, although it outperforms it somewhat.

Detection of COVID-19 Based on Deep Learning Methods: A Critical …

45

Fig. 5 Block Diagram of COVID-19 identification Based on chest X-ray images

Fig. 6 a Xception Net Training and Testing Loss with Consecutive Epochs. b Xception Net Accuracy Training and Testing with Consecutive Epochs

46

J. C. Babu et al.

Fig. 7 a Data from the Model of Xception Confusion Matrix train. b Uncertainty Matrix of the Xception model’s test data

Limitations 1. It doesn’t provide you any 3D data. 2. As the radiation is absorbed, the bones might obstruct important diagnostic data. 3. Higher elements do not have a strong interaction with them. The authors of [3], proposed a concept of Deep Learning for COVID-19 Detection based on CT images [12]. The pseudo code for calibrating the Convolutional Neural Network (CNN) and acquiring the exactness might found within the result as shown in Fig. 7. They calculated the slope, and the parameters of network were changed for each iteration by picking b CT pictures at random [13]. They confined the training steps rather than the iteration epoch, as opposed to the old conventional training approach. The authors utilised stochastic gradient descent (SGD) to select hyper parameters, The learning rate was programmed to 0.003, the force was programmed to 0.9, and the proportions of batch was programmed to 64. 512 × 512 × 512 × 512 image was the concluding input to the recommended structure, following which RGB reordering was applied [14]. The approach displays pseudo code for the Convolutional Neural Network is being fine-tuned (CNN) and achieving accuracy as shown in Fig. 8. Results In this part, we look at how well the replica performed in testing of COVID-19. Test Performance To train the models, we used the training parameters indicated in the section Hyper parameter settings for training. The findings are presented and compared to the most modern methods currently available [15]. In our lab, we employ the Random and BitM models, and they relate to random usage and pre-training approaches on ImageNet − 21k and ILSVRC-2012, which will be discussed in the section on the influence of

Detection of COVID-19 Based on Deep Learning Methods: A Critical …

47

Fig. 8 Block diagram of COVID-19 observation based on CT Scanning

parameter initialization. Our model was compared to the COVID-Net CT-2L, which is the most sophisticated [16]. Limitations 1. The price is exorbitant. 2. Radiation dosage is high. The authors of [17], proposed the work of Deep-Learning based detection of COVID19 using lung ultrasound Images. B-lines in LUS are frequently used to suggest the tissue of Fibrous, erythrogenic cells, or Thermal value of lung fluids may be present. These traits can be found in Interstitial lung disorders, pneumonia, and pulmonary emphysema are all examples of ARDS, to mention a few. It’s worth noting that isolated B-lines could be harmless (less than three). Its presence, however, may be associated to a specific lung ailment, based on the number of them and the distance between them. In general, the less lung aeration, the closer the B-Lines are together, the better (or even merged B-lines). A pleural effusion is a fluid collection in the pleural space with no air [18–20]. Ultrasound identifies not only effusion, as well as offers particulars about the type and identifies location where thoracentesis should be performed (i.e., draining the pleural area of fluid or air). Pneumonia and pulmonary edoema are both linked to pleural effusion. Furthermore, lung points with exclusive A-Lines are significant predictor of pneumothorax diagnosis. Lung consolidation is characterized by a thick tissue formation as well as the appearance of the white spots with aspiratory reinforcement. Because cancerous tissue or pleural effusion fills air spaces, this means that lung aeration is completely lost. Lung consolidations can be seen in a variety of conditions, including pneumonia, atelectasis, lung contusion, and ARDS. Results The CNN based replica were trained as well as verified with the help of various partitioning of the dataset for every5-fold cross demonstration experiment that was

48

J. C. Babu et al.

Fig. 9 Block diagram of COVID-19 identification using the image of lung ultrasound

conducted 5 times, resulting in matrices, (Loss/Exactness versus Era figure), ROC curves and 25 learning curves. In the accompanying material, you’ll get detailed 5 × fivefold cross validation findings as shown in Fig. 9. The average 55-fold cross validation results achieved in our experiment the various CNN-based replica. The Inception V3-based model had higher BACC, AUC-ROC and ACC mean values of 89.3, 97.3 and 89.1% Correspondingly. The Inception 3-based replica, on the other hand, offered the best mean metric values regarding accuracy, recall, and F1-score. The ResNet50-based model, on the other hand, had the lowest mean accuracy, recollect, and F1-score. Limitations 1. To produce a fine-tuned version of the examined procedures, a bigger dataset would be necessary. 2. Deep and intrapulmonary lesions are difficult to detect with ultrasound. The authors of [19], proposed the concept of FCOD: Fast COVID-19 Detector based on deep learning techniques. Compared to the fundamental designs, this technique uses fewer layers and filters. In addition, instead of convolutional layers, we used depth wise separable convolution layers. Three pieces make up the model. The model’s input layer was created using the chest X-ray pictures in the first stage. The max-pooling layer come after each of the depth wise divisible convolution layers after passing through four depth wise layers of divisible convolution. The Centre section was combined using 12 depth-separable convolution layers in the middle. Batch normalization follows the first and second portions of the process. Furthermore, batch normalizing allows for considerable gradients, as well as the outcomes are seen in the initial convergence. Following 16 layers of separable convolution and a drop out ratio of 0.2, the drop out layers are applied.

Detection of COVID-19 Based on Deep Learning Methods: A Critical …

49

Results Here, we include multiple assessment measures to signify the performance of the advanced technique. The replica is written in the Python, with the Keraschasis 2.2.4 as the backend and Tensorflow 1.14.0 as the frontend, and the hardware is provided by Google for free. Figure 10 depicts the suggested model’s confusion matrix and receiver operating characteristics (ROC) curve. Limitations 1. No 3D information is provided. 2. Significant diagnostic data might be obstructive due to the presence of bones. The researchers of [6] suggested a method called as Fast automated detection of COVID-19 from medical images using convolutional neural networks. The categorization framework for COVID-19 identification. B. Regression framework for analysing correlations betwixt lesion regions and the clinical markers. C. The categorization framework’s work flow for COVID-19 identification. Based on expert validation utilizing multi-model details from popular datasets and the youan hospital, a CNN-based classification framework performed exceptionally well. Figure 10 depicts the framework’s structure, which includes the two stages sub-frameworks. The main particulars of framework for common utilize scenarios are Q, L, M, and N. Results Data Set Properties Multi-model details from the various sources were utilized in the study X-data. CT details and clinical details utilized in the study were gathered from popular details programmed and one frontline named hospital details (Youan hospital). Every data set was splitted into two parts, test part and train-Val part ad test part utilizing a train-test-split function (TTSF) of scikit-learn library. The "Methods" section goes into the intricacies of the multi-model data types.

3 Conclusion As previously stated, we introduced various methods for Detection of Corona Virus in 2019 based on Deep Learning Techniques, such as discovery and diagnosis at an early stage of virus using Deep Learning Methods with the Fewest Complications as well as Costs, which are fundamental steps in anticipating pandemic steps and development. COVID-19 ON X-ray images of chestis detected as well as analyzed using Deep Learning. We used different CNN models to categorize COVID-19 patients gleaned from their chest X-ray images in this research. COVID-19 identification from CT

50

J. C. Babu et al.

Fig. 10 ROC curves and confusion matrix

Images based on Deep Learning is the third way. In this paper [20], he uses CT pictures to transfer learning COVID-19 testing, indicating that our replica, image Net21k was used to train it, is very generalizable when it comes to CT images. COVID19 detection utilizing lung ultrasound images is the fourth way. The employment of automated image diagnostic tools in this manner might significantly reduce the strain placed on health systems with a restricted number of expert practitioners.

Detection of COVID-19 Based on Deep Learning Methods: A Critical …

51

FCOD is a deep learning-based approach, is the fifth method. COVID-chest Xraydataset, a publicly available dataset, was used in this study. The sixth option is to use convolutional neural networks to quickly detect COVID-19 from medical photos. In this paper, we used DL-techniques to develop a computer-aided diagnostic approach for detecting virus in medical pictures. Finally, we compiled approaches for detecting COVID-19 utilizing deep learning methods and a variety of methodologies in one frame.

References 1. Zhou, P., Yang, X. L., Wang, X. G., Hu, B. et al. (2020). A pneumonia outbreak associated with a new coronavirus of probable bat origin. Nature. 2. Nagaraju, C. H., & Kondagandla, R. (2022). IoT based live monitoring public transportation security system by using Raspberry Pi, GSM& GPS. In V. K. Gunjan & J. M. Zurada (Eds.), Modern Approaches in Machine Learning & Cognitive Science: A Walkthrough. Studies in Computational Intelligence (Vol. 1027). Cham: Springer. https://doi.org/10.1007/978-3-03096634-8_43. 3. Rudra Kumar, M., Pathak, R., & Gunjan, V. K. (2022). Diagnosis and medicine prediction for COVID-19 using machine learning approach. In A. Kumar, J. M. Zurada, V. K. Gunjan & R. Balasubramanian (Eds.), Computational Intelligence in Machine Learning. Lecture Notes in Electrical Engineering (Vol. 834). Singapore: Springer. https://doi.org/10.1007/978-981-168484-5_10. 4. Katal, A. (2023). Leveraging fog computing for healthcare. In Deep Learning Technologies for the Sustainable Development Goals: Issues and Solutions in the Post-COVID Era (pp. 51–68). Singapore: Springer Nature Singapore. 5. Karimullah, S., & Vishnu Vardhan, D. (2022). Pin density technique for congestion estimation and reduction of optimized design during placement and routing. Applied Nanoscience. 6. Abdurrahman, Z., Li, M., & Wang, X. (2020). Comparative review of SARS-CoV-2, SARSCoV, MERS-CoV, and influenza a respiratory viruses. Frontiers Immunology, 11, 552909. 7. Siddiquee, K. N.-E.-A., Shabiul Islam, M., Singh, N., Gunjan, V. K, Yong, W. H., Huda, M. N., & Bhupal Naik, D. S. (2022) Development of algorithms for an IoT-Based smart agriculture monitoring system. Wireless Communications and Mobile Computing, 2022, Article ID 7372053, 16 p. https://doi.org/10.1155/2022/7372053. 8. Jaya Krishna, N., Shaik, F., Harish Kumar, G. C. V., Naveen Kumar Reddy, D., & Obulesu, M. B. (2021). Retinal vessel tracking using gaussian and radon methods. In A. Kumar & S. Mozar (Eds.) ICCCE 2020. Lecture Notes in Electrical Engineering (Vol. 698). Singapore: Springer. https://doi.org/10.1007/978-981-15-7961-5_37. 9. Vyas, S., Verma, S. S., & Prasad, A. (2023). Study of UAV management using cloud-based systems. In Deep Learning Technologies for the Sustainable Development Goals: Issues and Solutions in the Post-COVID Era (pp. 97–110). Singapore: Springer Nature Singapore. 10. Gunjan, V. K., Shaik, F., & Kashyap, A. (2021). Detection and analysis of pulmonary TB using bounding box and K-means algorithm. In A. Kumar & S. Mozar (Eds.), ICCCE 2020. Lecture Notes in Electrical Engineering (Vol. 698). Singapore: Springer. https://doi.org/10.1007/978981-15-7961-5_142. 11. Usman, M., Gunjan, V. K., Wajid, M., & Zubair, M. (2022). Speech as a biomarker for COVID19 detection using machine learning. Computational Intelligence and Neuroscience, 2022. 12. Funk, C. D., Laferrière, C., & Ardakani, A. (2020). A snapshot of the global race for vaccines targeting SARS-CoV-2 and the COVID-19 pandemic. Frontiers Pharmacology, 11, 937.

52

J. C. Babu et al.

13. Karimullah, S., Vishnu Vardhan, D., & Javeed Basha, S. (2020). Floorplanning for placement of modulesin VLSI physical design using harmonysearch technique. ICDSMLA 2019, Lecture Notes in Electrical Engineering (Vol. 601). Springer Nature Singapore Pte Ltd. 14. Lu, R., Zhao, X., Li, J., Niu, P. B. et al., (2020) Tan Genomic characterization and epidemiology of 2019 novel corona virus: implications for virus origins and receptor binding Lancet (Vol. 395, pp. 565–574). 15. Sunitha, P., Ahmad, N., Barbhuiya, R. K., Gunjan, V. K., & Ansari, M. D. (2022, May). Impact of covid-19 on education. In ICCCE 2021: Proceedings of the 4th International Conference on Communications and Cyber Physical Engineering (pp. 1191–1197). Singapore: Springer Nature Singapore. 16. Chen, Z., Lv, N., Liu, P., Fang, Y., Chen, K., & Pan, W. (2020). Intrusion detection for wireless edge networks based on federated learning. IEEE Access, 8, 217463–217472. 17. Krishna, S. L. V., Abdul Rahim, B., Shaik, F., & Soundara Rajan, K. (2010). Lossless embedding using pixel differences and histogram shifting technique. In Recent Advances in Space Technology Services and Climate Change 2010 (RSTS & CC-2010), Chennai, India (pp. 213-216). https://doi.org/10.1109/RSTSCC.2010.5712850. 18. Verma, A. (2023). ANN: Concept and application in brain tumor segmentation. In Deep Learning Technologies for the Sustainable Development Goals: Issues and Solutions in the Post-COVID Era (pp. 175–187). Singapore: Springer Nature Singapore. 19. Wu, J., Liu, J., Li, S., Peng, Z. et al., Detection and analysis of nucleic acid in various biological samples of COVID-19 patients. Tropical Medicine and Infectious Disease, 2020. 20. Verma, A. (2023). Automation of brain tumor segmentation using deep learning. In Deep Learning Technologies for the Sustainable Development Goals: Issues and Solutions in the Post-COVID Era (pp. 189–202). Singapore: Springer Nature Singapore. 21. Dveksler, G. S., Pensiero, M. N., et al. (1991). Cloning o the mouse hepatitis virus (MHV) receptor: Expression in human and hamster cell lines confers susceptibility to MHV. Journalof Virology, 65(12), 6881–6891.

Performance Evaluation of GA, HS, PSO Algorithms for Optimizing Area, Wirelength Using MCNC Architectures Shaik Karimullah, D. Vishnuvardhan, Vinit Kumar Gunjan, and Fahimuddin Shaik

1 Introduction Gordon Moore, who helped start Intel, found out in the 1960s that the number of transistors on silicon ICs had grown every year since they were first made. Moore’s law shows that this trend is likely to keep going strong in the near future [1]. Even though growth has slowed, the number of devices per square unit has gone up about every 18 months since then. Moore’s law describes a type of growth called exponential growth, which is likely to keep going on forever. Due to short channel effects in transistors, some studies say that the physical limit of integration could be reached as early as 2017. Integrated circuits have changed a lot over the past 50 years, with the number of devices on a single silicon chip growing from a few thousand to more than a million. The IC industry for electronic devices like cell phones has made a lot of progress, especially in the last ten years [2]. This means that in the next few years, chips with a higher number of transistors that can run at different Giga Hertz rates will be possible to make. Such chips could start a new trend in electronic devices, making it possible for cool things like augmented reality, wearable computers, and computers that can be put into the body. It could make low-cost point-to-point communication around the world possible for everyone [3, 4]. As VLSI technology improved quickly, it became possible to automate some of the steps needed to design, test, and make S. Karimullah (B) · F. Shaik Department of ECE, AITS, Rajampet, India e-mail: [email protected] D. Vishnuvardhan Department of ECE, JNTUACE, Ananthapuramu, India e-mail: [email protected] V. K. Gunjan Department of Computer Science and Engineering, CMR Institute of Technology, Hyderabad, India © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 V. K. Gunjan et al. (eds.), Modern Approaches in Machine Learning and Cognitive Science: A Walkthrough, Studies in Computational Intelligence 1117, https://doi.org/10.1007/978-3-031-43009-1_5

53

54

S. Karimullah et al.

VLSI-specific circuits. In today’s technological society, integrated circuits are made up of many different large electronic parts. These parts are put together by putting them in a certain order on a single piece of silicon, called a “wafer,” as shown in Fig. 1. Layout is the process by which an IC designer changes a description of a circuit into a description of its shape. There are several layers of flat geometric shapes that make up a layout. The layout is checked to make sure it meets all the design standards that are wanted. The next step is to make a set of design files that describe the layout of the design. Pattern generator files can be made by the optical pattern generator (design files) [5]. The files that are made are then used to make “Masks,” which are patterns for laying out circuits. During the manufacturing process, these masks are used to line up the silicon case using the different stages of Photo-Lithography [6, 7]. For the device to be made, the minimum size of the geometric patterns’ features and the minimum distance between them must be known with accuracy.

2 Literature Review The authors of [8] came up with a new method that changes integer coding based on the number of modules. In this paper, mutation and crossover operators from genetic algorithms are used with discrete particle swarm optimization to come up with better solutions. The authors compare and contrast Simulated Annealing with B*Tree, Particle Swarm, and DPSO algorithms. In the trials, the MCNC and GSRC benchmark circuits were used, and the proposed method did well in terms of placement because it avoided local minimums. The researchers in [9], K. Premalatha and A.M. Natarajan made GA a hybrid and looked into how it could be used with other optimization techniques. This article talked about hybrid methods that use both Particle Swarm Optimization (PSO) and Genetic Algorithms (GA). Both PSO and GA are heuristic and population-based search strategies that have been used to solve optimization problems that are getting harder and harder to solve. In order to get rid of the early stagnation seen in traditional PSO, this article gives you the change tactics for PSO that use GA. The authors of [10] showed comparison results for non-slicing hard module VLSI floor layout using B*Tree representation with a hybrid genetic algorithm and memetic algorithm. The results come from MCNC benchmark circuits that have HGA. The results show that a hybrid genetic algorithm could be used to cut down on the size of the circuit. Then, in 2012 [11], the authors came up with a work called “Optimization of Floor-planning Using Genetic Algorithm”. The work shown here is a creative way to solve problems with VLSI floor planning. This method is based on the POEMS algorithm for optimising repeating prototypes with improving changes [12–14]. In each cycle, a Genetic Algorithm (GA) is used for local search because both methods have been useful in the past for solving problems that are similar. GA was used to do well-known MCNC benchmark tasks and was evaluated on those tasks. Experiments show that the Genetic Algorithm quickly comes up with the best answers to all benchmark problems studied. The authors looked at a lot of algorithms in 2013.

Performance Evaluation of GA, HS, PSO Algorithms for Optimizing …

55

They found that while not all of them were good, a few were very good and have become popular tools for solving real-world and high-definition problems [15]. Some algorithms are not studied enough because of how they work [16–19]. The goal of this review is to make a complete list of all optimization algorithms that have been written about so that they can inspire more research [20]. In 2014, the researchers in [21] proposed a way to improve VLSI physical design automation. The authors used a Hybrid Genetic algorithm to find the answer. The people who wrote this article did all the physical design calculations on their own. A Genetic Algorithm is used to improve things on a global scale [22]. Simulated annealing is used to improve things locally. Then, in 2015, Debao Chen, Jing Chen, and others came up with a new way to improve the global performance of particle swarm optimization (PSO) by extending the domain exploration of the optimal solution in the current generation and the optimal solution reached by each particle [23–25]. The algorithm is easy to use because it doesn’t change the core of PSO, and it can be easily added to other optimization methods to speed things up.

3 Simulation Results During the Placement and Routing of Computational Logic Blocks with respect to MCNC Benchmarks the significant parameters such as Number of Blocks, Selection of Corners, Number of Iterations, Best Placement, and Routing Wirelength have to be correlated with respect to the parameters of HS, PSO, GA algorithms as shown in Fig. 1. PARAMETERS OF MCNC BENCHMARKS Number of Blocks Number of Corners Iterations Placement Wirelength

Harmony States HMCR Rehearsals Best Result PAR HS

Particles Particle validation Velocity updation Convergence Evolution Time PSO

Chromosomes Selection Mutations Best Chromosome Crossover duration GA

Fig. 1 Analogies between optimization algorithm parameters and MCNC Benchmarks

56

S. Karimullah et al.

3.1 Simulation Results of MCNC Benchmark Circuits for PSO Algorithm For the MCNC APTE Benchmark circuit, which has 9 computational blocks, 97 nets, and a standard area of 46.5616 mm2 , the Particle Swarm Optimization algorithm is used. Figure 2 shows how the simulation turned out. So that the best places to put the blocks can be found, they are moved around so that they don’t overlap. The PSO algorithm has better criteria for exploratory search that lead to the best optimal result. The plot shows how the blocks are set up based on the internal properties of the APTE circuit being looked at, which is a part of the MCNC benchmark. Here, the best placement position is found by looking at where the object is now and how much space it needs. The Particle Swarm Optimization method uses 1000 iterations. At each iteration, the position of the Computation Logic Block is changed though the presentbestcost is better compared to the preceding best cost. If not, the position stays the same. The MCNC XEROX Benchmark circuit, which has 10 computational blocks, 203 nets, and a standard area of 19.3503 mm2 , uses the Particle Swarm Optimization algorithm. Figure 3 shows what happened in the simulation. To get a better arrangement without losing any information, the overlap needs to be stopped. So that there are no overlaps, the blocks are shuffled to find the best places to put them. The best places to put things are arranged by stack. Here, the best place to put something is decided by comparing where it is now to where there is the least amount of space. The Particle Swarm Optimization method needs 1000 iterations, and at each iteration, the placement of the Computation Logic Block is changed if the existing best cost is lower compared with the earlier best cost. If not, position stays the same.

Fig. 2 Simulation result with PSO for MCNC APTE circuit

Performance Evaluation of GA, HS, PSO Algorithms for Optimizing …

57

Fig. 3 Simulation result with PSO for MCNC XEROX circuit

The Particle Swarm Optimization method is used for the MCNC Benchmark HP circuit, which has 11 computational blocks, 83 nets, and a standard size of 8.8306 mm2 . Figure 4 shows how the simulation turned out. So that there are no overlaps, the blocks are moved around to find the best way to put them together. The gaps between the blocks can be bigger than they need to be, which shows how limited the process is in this case. Here, the best place to put something is found by comparing where it is now to where there is the least amount of space.

Fig. 4 Simulation result with PSO for MCNC HP circuit

58

S. Karimullah et al.

Fig. 5 Simulation result with PSO for MCNC AMI33 circuit

The Particle Swarm Optimization method needs 1000 iterations, and at each iteration, the Computation Logic Block’s placement location is changed then the contemporary best cost is fewer than earlier best cost. Otherwise, position stays the same. The MCNC AMI33 Benchmark circuit, which has 33 computational blocks, 123 nets, and a standard area of 1.1564 mm2 , uses the Particle Swarm Optimization algorithm. Figure 5 shows how the simulation turned out. In order to discover the best placement sites, the blocks are shuffled so that there is no overlap. To get a better arrangement without losing any information, the overlap needs to be stopped. After looking at where things already are and how little space there is, the best place to put something is chosen. It can be seen that the spaces between the blocks are bigger than they need to be. This shows how limited the process is. The Particle Swarm Optimization method needs 1000 iterations. Best cost now is a lesser amount compare with best cost before, the placement location of the Computation Logic Block is changed. If not, the position stays the same. The MCNC AMI 49 Benchmark circuit uses the Particle Swarm Optimization method. This circuit has 49 computational blocks, 408 nets, and a standard size of 35.4454 mm2 . Figure 6 shows how the simulation turned out. In order to find the best places to put the blocks, they are moved around so that they don’t touch each other. After looking at where things already are and how little space there is, the best place to put something is chosen. The Particle Swarm Optimization method needs 1000 iterations. The best cost now is less than the best cost before, the placement location of the Computation Logic Block is changed. If not, the position stays the same.

Performance Evaluation of GA, HS, PSO Algorithms for Optimizing …

59

Fig. 6 Simulation result with PSO for MCNC AMI49 circuit

3.2 Simulation Results of MCNC Benchmark Circuits for Genetic Algorithm For the MCNC APTE Benchmark circuit, which contains 9 computational blocks, 97 nets, and a standard size of 46.5616 mm2 , the Genetic algorithm is used. Figure 7 shows what happened in the simulation. To find the best places to put the blocks, they are moved around so that they don’t touch each other. After looking at the current position and how little space can be used, the best place to put something is found. Because of how GA works, putting blocks next to each other with space between them leads to bad results. The Genetic Algorithm method needs 1000 iterations. If the best cost now is less than the best cost before, the placement of the Computation Logic Block is changed. If the best cost now is the same as the best cost before, the position stays the same. The MCNC XEROX Benchmark circuit, which has ten computational blocks, 203 nets, and a standard area of 19.3503 mm2 , is optimised using the Genetic algorithm. Figure 8 shows what the simulation came up with. To determine the optimal placement of the blocks, they are shuffled in such a way that no blocks overlap. Here, the optimal placement position is determined by comparing the present position to the smallest available space. The spaces between the blocks can be found to be wider than they need to be, indicating the process’s limitations. These gaps must be reduced in size in order to optimize the area and wirelength. The Genetic Algorithm process is repeated 1000 times, and during each iteration, if the current best cost is less than the prior best cost, the placement position of the Computation Logic Block is altered; otherwise, the position remains unchanged.

60

S. Karimullah et al.

Fig. 7 Simulation result with GA for MCNC APTE circuit

Fig. 8 Simulation result with GA for MCNC XEROX circuit

The MCNC Benchmark HP circuit is made with the Genetic algorithm. It has 11 computational blocks, 83 nets, and a standard area of 8.8306 mm2 . Figure 9 shows what the simulation came up with. So that there are no overlaps, the blocks are shuffled to find the best places to put them. Here, the best place to put something is decided by comparing where it is now to where there is the least amount of space.

Performance Evaluation of GA, HS, PSO Algorithms for Optimizing …

61

Fig. 9 Simulation result with GA for MCNC HP circuit

To avoid overlaps, the best care must be taken to find the available space. The Genetic Algorithm method needs 1000 iterations, and at each iteration, the placement of the Computation Logic Block is changed then recent best cost is lesser than the prior best cost. If not, the position stays the same. The GA-Genetic Algorithm is used for MCNC AMI33 Benchmark Circuit, which has 33 computational blocks, 123 nets, and a standard area of 1.1564 mm2 . Figure 10 shows what happened in the simulation. To find the best places to put the blocks, they are moved around so that they don’t touch each other. After looking at the current position and how little space can be used, the best place to put something is found. It’s possible that the spaces between the blocks are bigger than they need to be, which shows how limited the process is. As the GA is hard to compute, putting the blocks in order can take a long time, which is a clear limitation. To get the best area and wire length, these gaps need to be made smaller. The Genetic Algorithm Optimization method needs 1000 iterations. If the best cost now is less than the best cost before, the placement of the Computation Logic Block is changed. If the best cost now is the same as the best cost before, the position stays the same. The MCNC AMI 49 Benchmark circuit, which has 49 computational blocks, 408 nets, and a standard area of 35.4454 mm2 , is optimised with a genetic algorithm. Figure 11 shows what the simulation came up with. To figure out where the blocks should go best, they are mixed up so that none of them overlap. Here, the best place to put something is found by comparing where it is now to the smallest space available. But because the structure of Genetic Algorithms is more complicated, they don’t scale as well. The Genetic Algorithm process is repeated 1000 times. During each iteration, the placement position of the Computation Logic Block is changed if the current best cost is less than the previous best cost. Otherwise, the position stays the same. Since GA is based on chance, not all iterations lead to good results.

62

S. Karimullah et al.

Fig. 10 Simulation result with GA for MCNC AMI33 circuit

Fig. 11 Simulation result with GA for MCNC AMI49 circuit

3.3 Simulation Results of MCNC Benchmark Circuits for HS Algorithm The Harmony Search algorithm is used for the MCNC APTE Benchmark circuit, which consists of 9 computational blocks, 97 nets, and a standard area of 46.5616 mm2 . The simulation result is shown in Fig. 12. To find the optimal placements for laying the blocks, they are shuffled so that there is no overlap. Here, the ideal placement

Performance Evaluation of GA, HS, PSO Algorithms for Optimizing …

63

Fig. 12 Simulation result with HS for MCNC APTE circuit

location is determined by comparing the present position to the smallest amount of space available. The Harmony Search Optimization method requires 1000 iterations, and for each iteration, if the current best cost is lower than the previous best cost, the placement location of the Computation Logic Block is altered; otherwise, the position remains unchanged. The Particle Harmony Search algorithm is used in the MCNC XEROX Benchmark circuit. It is made up of 10 computational blocks, 203 nets, and a standard area of 19.3503 mm2 . The result of the simulation is shown in Fig. 13. When the blocks are moved around to find the best place to put them, they don’t touch each other. In general, the HS algorithm imitates the musician’s height by giving the best nodes for music, which leads to the best arrangement of blocks. Here, the best place to put something is decided by comparing where it is now to the smallest amount of space possible. It’s possible that the spaces between the blocks are too big, which shows how limited the process is. These spaces need to be smaller so that the most space and wire length can be used. In the Harmony Search Optimization method, there are 1000 iterations, and the placement of the Computation Logic Block is changed if the best cost now is better than the best cost before. Other than that, it stays the same. The Harmony Search algorithm is used in the MCNC Benchmark HP circuit, which contains 11 computational blocks, 83 nets, and a standard area of 8.8306 mm2 . Figure 14 shows what the simulation came up with. So that there are no overlaps, the blocks are shuffled to find the best places to put them. Here, the best place to put something is decided by comparing where it is now to where there is the least amount of space. The HS algorithm tends to always give a perfect state of harmony,

64

S. Karimullah et al.

Fig. 13 Simulation result with HS for MCNC XEROX circuit

but it is easy for it to get stuck in local optima, which slows convergence and makes the arrangement messier. The Harmony Search Optimization method needs 1000 iterations, and at each iteration, the placement of the Computation Logic Block is changed if the current best cost is lower than the previous best cost. If not, the position stays the same.

Fig. 14 Simulation result with HS for MCNC HP circuit

Performance Evaluation of GA, HS, PSO Algorithms for Optimizing …

65

Fig. 15 Simulation result with HS for MCNC AMI33 circuit

The Harmony Search method is used on the MCNC AMI33 Benchmark circuit, which has 33 computational blocks, 123 nets, and a standard area of 1.1564 mm2 . Figure 15 shows what happened in the simulation. To find the best places to put the blocks, they are moved around so that they don’t touch each other. After looking at the current position and how little space can be used, the best place to put something is found. It’s possible that the spaces between the blocks are bigger than they should be, showing that there are problems with the process. The HS method usually leads to a perfect state of harmony, but it can get stuck in local optima, which slows convergence and makes the arrangement look messier. To make the best use of the space and wire length, these gaps need to be made smaller. The Harmony Search Optimization method needs 1000 iterations. If the current best cost is less than the previous best cost, the position of the Computation Logic Block is changed. If the current best cost is the same as the previous best cost, the position of the Computation Logic Block stays the same. For the MCNC AMI49 Benchmark circuit, which has 49 computational blocks, 408 nets, and a standard area of 35.4454 mm2 , the Harmony Search algorithm is used. Figure 16 shows what the simulation came up with. So that there are no overlaps, the blocks are shuffled to find the best places to put them. Here, the best place to put something is decided by comparing where it is now to where there is the least amount of space. It’s possible that the spaces between the blocks are bigger than they need to be, showing weaknesses. To get the best area and wire length, these gaps need to be made smaller. The HS method usually leads to a perfect state of harmony, but it can get stuck in a local optimum, which slows convergence and makes the arrangement look messier. The Harmony Search Optimization method needs 1000 iterations, and at each iteration, the placement of the Computation Logic Block is changed if the current best cost is lower than the previous best cost. If not, the position stays the same.

66

S. Karimullah et al.

Fig. 16 Simulation result with HS for MCNC AMI49 circuit

4 Statistical Comparison Table 2 shows how the evolutionary algorithms were used to arrange the blocks so that they met the requirements for the MCNC benchmark circuit in terms of area, wire length, and dead space. This table compares the area that was found to the standard values in Tables 1, 3, and 4. It can be seen that the values for Area, Wirelength, and Dead space are closer to what would be expected. In this case, the number of blocks and standard MCNC Benchmarks are used to figure out the area, wirelength, and dead space. “Area, Wirelength, and Dead space” is a type of objective function that sums up the whole performance into a single important parameter so that the best design solutions can be found. For each Benchmark circuit, the numbers for Area, Wirelength, and Deads pace show how well the solution fits the given number of cells and nets. It’s important to note that XEROX BMC has the largest Area, Wirelength, and Dead space in GA. This means that the solution comes closest to meeting the overall needs Table 1 Characteristics of MCNC benchmark circuits

MCNC circuit No. of nets No. of cells Standard area (mm2 ) AMI33

123

33

1.1

XEROX

203

10

19.3

AMI49

408

49

35.4

HP

83

11

08.8

APTE

97

9

46.5

Performance Evaluation of GA, HS, PSO Algorithms for Optimizing …

67

Table 2 Comparison of GA values with standard values S. no

MCNC benchmark

No. of cells

Standard area (mm2 )

Obtained area (mm2 )

Wire length (mm)

Dead space (%)

1

APTE

9

46.56

48.95

433

4.88

2

HP

11

08.83

11.55

184

23.54

3

XEROX

10

19.35

22.95

989

15.68

68.6

4477

48.33

198

38.50

4

AMI49

49

35.44

5

AMI33

33

1.15

1.87

Table 3 Comparison of HS values with standard values S. no

Benchmark circuit

1

APTE

2

XEROX

No. of cells

Standard area (mm2 )

Obtained area (mm2 )

Wire length (mm)

Dead space (%)

9

46.56

48.43

385

3.86

10

19.35

21.67

905

10.70

3

HP

11

08.83

10.96

146

19.43

4

AMI33

33

1.15

1.82

193

36.81

5

AMI49

49

35.44

66.49

4236

46.69

Wire length (mm)

Dead space (%)

Table 4 Comparison of PSO values with standard values S. no

Benchmark circuit

1

APTE

No. of cells 9

Standard area (mm2 )

Obtained area (mm2 )

46.56

48.21

376

3.42

2

XEROX

10

19.35

20.96

897

7.68

3

HP

11

08.83

10.82

134

18.39

4

AMI33

33

1.15

1.66

190

30.72

5

AMI49

49

35.44

62.32

3377

43.13

of the intended solution. The GA Algorithm was used to place blocks on MCNC benchmark circuits. Table 2 shows the area and wirelength values that were found. The Harmony Search Algorithm for Block Placement was also used to analyze the MCNC Benchmark circuits; the area and wire length, as well as the area, wire length, and dead space, are listed in Table 3. As can be seen, the resultant area is optimal in comparison to the standard values for the respective benchmark circuits. Here, the values for Area, Wirelength, and Dead space are better for APTE and HP, indicating the fit element of the optimal solution in solving the Block Placement Problem of the task. When dealing with Area, Wirelength, and Dead space values, it is necessary to keep in mind that there is no hard and fast rule dictating that a certain value of the concerned parameters, such as Area and wirelength, must be used. Furthermore, the PSO Algorithm was explored for block placement on MCNC benchmark circuits, and

68

S. Karimullah et al.

the acquired area and related wirelength values are shown in Table 4. When compared to standard values, the produced area has optimal values. Area, Wirelength, and Dead space all prove to be optimal values for the HP benchmark circuit at 1.025, with an area increase of 18%.PSO arranges the particles randomly by promoting the values acquired for Area, Wirelength, and Dead space in relation to the best value for the individual particle and the best value for the swarm’s position, respectively, in order to achieve an ideal solution.

5 Conclusion As the simulation results have been initiated for Placement and Routing with six Optimization Algorithms, as three algorithms namely Genetic Algorithm, Particle Swarm Optimization and Harmony Search Algorithm have provided best cost solution along with less computation time. These three algorithms have been investigated on the standard MCNC Benchmark circuits have been compared in terms of Obtained Area, Wirelength and Dead space. As per theoretical and standard norms the obtained area must be less and nearer to standard area values as provided in Table 1. It is obvious that Particle Swarm Optimization and Harmony Search algorithms have yielded the minimum area when compared to Genetic Algorithm. Similarly the significant parameters Wirelength and Dead space were also computed through the three Optimization algorithms and it is observed that Particle Swarm Optimization and Harmony Search algorithms have provided the best and Optimum values.

References 1. Gao, X., Jiang, Y. M., Shao, L., Raspopovic, P.; Verbeek, M. E., Sharma, M., Rashingkar, V., & Jalota, A. (2022). Congestion and timing aware macro placement using machine learning predictions from different data sources: cross-design model applicability and the discerning ensemble. In Proceedings of the ISPD ’22, 2022 International Symposium on Physical Design, Virtual, March 27–30, 2022 (pp. 195–202). 2. Karimullah, S., Sai Sumanth Goud, E., & Lava Kumar Reddy, K. (2023). Spectral efficiency for multi-bit and blind medium estimation of DCO-OFDM used vehicular visible light communication. In: A. Kumar, S. Senatore & V. K. Gunjan (Eds.), ICDSMLA 2021. Lecture Notes in Electrical Engineering (Vol. 947). Singapore: Springer. https://doi.org/10.1007/978-981-195936-3_83. 3. Krishna, S. L. V., Abdul Rahim, B., Shaik, F., & Soundara Rajan, K. (2010) Lossless embedding using pixel differences and histogram shifting technique. Recent Advances in Space Technology Services and Climate Change 2010 (RSTS & CC-2010), Chennai, India (pp. 213–216). https:// doi.org/10.1109/RSTSCC.2010.5712850. 4. Kavitha, A., et al. (2022). Security in IoT mesh networks based on trust similarity. IEEE Access, 10, 121712–121724. https://doi.org/10.1109/ACCESS.2022.3220678 5. Ojha, N., Kumar, A., Tyagi, N., Ranjan, P., & Vaish, A. (2023). Use of machine learning in forensics and computer security. In V. Sarveshwaran, J.IZ. Chen, & D. Pelusi (Eds.), Artificial

Performance Evaluation of GA, HS, PSO Algorithms for Optimizing …

6.

7.

8.

9.

10.

11.

12.

13.

14.

15.

16.

17.

18.

19. 20.

69

Intelligence and Cyber Security in Industry 4.0. Advanced Technologies and Societal Change. Singapore: Springer. https://doi.org/10.1007/978-981-99-2115-7_9. Liang, R., Xiang, H., Pandey, D., Reddy, L., Ramji, S., Nam, G. J., & Hu, J. (2020). DRC hotspot prediction at sub-10 nm process nodes using customized convolutional network. In Proceedings of the ISPD’20, 2020 International Symposium on Physical Design, Taipei, Taiwan, September 20–23, 2020 (pp. 135–142). Nagaraju, C. H., Kondagandla, R. (2022). IoT based live monitoring public transportation security system by using raspberry Pi, GSM& GPS. In: V. K. Gunjan, J. M. Zurada, (Eds.), Modern Approaches in Machine Learning & Cognitive Science: A Walkthrough. Studies in Computational Intelligence (Vol. 1027). Cham: Springer. https://doi.org/10.1007/978-3-03096634-8_43. Prasad, P. S., Beena Bethel, G. N., Singh, N., Gunjan, V. K., Basir, S., & Miah, S. (2016). Blockchain-Based privacy access control mechanism and collaborative analysis for medical images. Security and Communication Networks, 2022, Article ID 9579611, 7 p. https://doi.org/ 10.1155/2022/9579611.ShivetanshTickoo and Kohli, S. (2016). A review on: floorplanning— based design methodology. International Journal of Advanced Research in Computer Science and Software Engineering, 6(6), 184–188. Hung, W. L., Xie, Y., Vijaykrishnan, N., Addo-Quaye, C., Theocharides, T., & Irwin, M. J. (2005). Thermal-Aware floorplanning using genetic algorithms. In Proceedings of 6th International Symposium on IEEE Quality of Electronic Design, March 21, 2005 (pp. 634–639). https://doi.org/10.1109/ISQED.2005.122. Rajagopal, M., & Ramkumar, S. (2023). Adopting Artificial Intelligence in ITIL for Information Security Management—Way Forward in Industry 4.0. In Artificial Intelligence and Cyber Security in Industry 4.0 (pp. 113–132). Singapore: Springer Nature Singapore. Sivayamini, L., Venkatesh, C., Fahimuddin, S., Thanusha, N., Shaheer, S., & Sree, P. S. (2017). A novel optimization for detection of foot ulcers on infrared images. In 2017 International Conference on Recent Trends in Electrical, Electronics and Computing Technologies (ICRTEECT), Warangal, India (pp. 41–43). https://doi.org/10.1109/ICRTEECT.2017.36. Karimullah, S., Vishnuvardhan, D. (2020). “Experimental analysis of optimization techniques for placement and routing in Asic design” ICDSMLA 2019. Lecture Notes in Electrical Engineering (Vol. 601). Springer Nature Singapore Pte Ltd. Nakatake, S., Furuya, M., & Kajitani, Y. (1998). Module placement on BSG-structure with preplaced modules and rectilinear modules. In Proceedings of Asia and South Pacific—Design Automation Conference, February 10, 1998 (pp. 571–576). Karimullah, S., Vishnu Vardhan, D., Basha, S. J. (2020). Floorplanning for placement of modules in VLSI physical design using harmony search technique, ICDSMLA 2019. Lecture Notes in Electrical Engineering (Vol. 601). Springer Nature Singapore Pte Ltd. Hong, X., Huang, G., Cai, Y., Gu, J., Dong, S., Cheng, C. K., & Gu, J. (2000). Corner block list: An effective and efficient topological representation of nonslicing floorplan. Proceedings of IEEE/ACM International Conference on Computer-Aided Design, 5, 8–12. Kumar, S., Ansari, M. D., Naik, M. V., Solanki, V. K., & Gunjan, V. K. (2020). A comparative case study on machine learning based multi-biometric systems. Advances in Cybernetics, Cognition, and Machine Learning for Communication Technologies, 353–365. Tabrizi, A. F., Darav, N. K., Rakai, L., Bustany, I., Kennings, A., & Behjat, L. E. (2020). Predictor: A deep learning framework to identify detailed routing short violations from a placed netlist. IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems, 39, 1177–1190. Kumar, R., & Sundaramurthy, S. (2023). AI and IoT in manufacturing and related security perspectives for industry 4.0. In Artificial Intelligence and Cyber Security in Industry 4.0 (pp. 47–70). Singapore: Springer Nature Singapore. Chen, J., Zhu, W., & Ali, M. (2011). A hybrid simulated annealing algorithm for nonslicing VLSI floorplanning. Systems, Man, and Cybernetics, 41, 544–553. Chan, W. T. J., Ho, P. H., Kahng, A. B., Saxena, P. R. (2017). Optimization for industrial designs at sub-14 nm process nodes using machine learning. In Proceedings of the 2017 ACM

70

21.

22.

23.

24.

25.

S. Karimullah et al. on International Symposium on Physical Design, Portland, OR, USA, 19–22 March, 2017 (pp. 15–21). Kumar, S., Gunjan, V. K., Ansari, M. D., & Pathak, R. (2022). Credit card fraud detection using support vector machine. In V. K. Gunjan, J. M. Zurada (Eds.), Proceedings of the 2nd International Conference on Recent Trends in Machine Learning, IoT, Smart Cities and Applications. Lecture Notes in Networks and Systems (Vol. 237). Singapore: Springer. https://doi. org/10.1007/978-981-16-6407-6_3. Karimullah, S., Basha, S. J., Guruvyshnavi, P., Sathish Kumar Reddy, K., & Navyatha, B. (2020). A genetic algorithm with fixed open approach for placements and routings ICCCE (pp. 599–610), Publisher Springer. Chen, J., Kuang, J., Zhao, G., Huang, D. J. H., & Young, E. F. Y. (2020). PROS: A plug-in for routability optimization applied in the state-of-the-art commercial EDA tool using deep learning. In Proceedings of the 39th International Conference on Computer-Aided Design, Virtual, November 2–5, 2020 (pp. 1–8). Das, P., & Saif, S. (2023). Intrusion detection in IoT-Based healthcare using ML and DL approaches: A case study. In Artificial Intelligence and Cyber Security in Industry 4.0 (pp. 271– 294). Singapore: Springer Nature Singapore. Karimullah, S., Vishnuvardhan, D., & Bhaskar, V. (2022). An improved harmony search approach for block placement for VLSI design automation. Wireless Personal Communications, 127, 3041–3059. https://doi.org/10.1007/s11277-022-09909-2

An Enhanced Woelfel Image Noise Filter K. Riyazuddin, Shaik Bajidvali, B. Abdul Raheem, Shaik Karimullah, N. Merrin Prasanna, and Pabbati Swathi

1 Introduction Image denoising is the removal of noise or distortions from an image [1]. There are numerous applications where blurred images can be made clear. As a result of medical imaging research, numerous diagnostic methods, including CT, MRI, and ultrasound, have been developed [2]. Each has a unique set of benefits and drawbacks [3]. The procedure of creating visual representations of the interior of the body for medical diagnosis is known as medical imaging [4]. It aids in the discovery of internal structures hidden beneath the skin and bones, as well as in the treatment and diagnosis of diseases [5]. It detects abnormalities using a database of physiology and normal anatomy [6]. Organs and tissues that can be imaged for medical purposes are removed [7]. This is a pathology procedure, not a medical imaging procedure [8]. Medical imaging is a subset of biological imaging, and image noise suppression is a major issue, particularly when images are obtained under adverse conditions [9].

K. Riyazuddin (B) · S. Karimullah · N. Merrin Prasanna · P. Swathi Annamacharya Institute of Technology and Sciences, Rajampet, India e-mail: [email protected] S. Bajidvali Department of Electronics and Communication Engineering, Narasaraopeta Engineering College, Narasaraopet, India B. A. Raheem Joginpally B.R. Engineering College, Hyderabad, India © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 V. K. Gunjan et al. (eds.), Modern Approaches in Machine Learning and Cognitive Science: A Walkthrough, Studies in Computational Intelligence 1117, https://doi.org/10.1007/978-3-031-43009-1_6

71

72

K. Riyazuddin et al.

1.1 Need and Significance of the Proposed Work Medical imaging and diagnostics techniques have grown in popularity over the last two decades as a result of rapid advancements in computing, internet, data storage, and Wireless technology [10]. The impact of these advancements can be seen in the field of medicine and medical sciences, which allows for more effective diagnosis and treatment of various diseases [11]. • Further more, medical imaging is frequently justified in the surveillance of a disease that has already been diagnosed and treated [12]. • The outcomes proved that suggested technique emerge significantly better in values of different quantitative measures such as signal to noise ratio, high snr will be required [13]. • It improves Image quality. The Proposed denoising method outperforms the existing method in experimental results [14]. • It improves medical image processing performance [15].

2 Proposed Methodology • The unique filter fabrication would be preferred almost universally because the complexity of the filter does not affect its realization in a digital system as shown in Fig. 1. • Both filters were identical in size (order) and took about the same amount of time to process a photo. • In this technique we are using Woelfel method (using two fir filters). • Whether decimating or interpolating, the use of FIR filters allows for the omission of some calculations, resulting in significant computational efficiency. • In contrast, if IIR filters are used, each output must be computed independently, regardless of whether it will be rejected (thus adding feedback into the filter). • They work well in applications requiring several rates. We suggest that "interpolation" refers to either increasing the sampling rate or both, whereas "decimation" refers to decreasing the sampling rate. • FIR filters are employed for decimating as well as interpolating, and they allow the elimination of certain calculations, considerably enhancing computer performance. • In contrast, even if a result is eliminated (to incorporate feedback), each output of IIR filters must be computed independently.

An Enhanced Woelfel Image Noise Filter Fig. 1 Block diagram for proposed methodology using woelfel filter

Noi

73

PSD of ima

COL MAP RGB GR

BRAIN MRI

BRAIN SCA DATAB

PSD ima

Noi Ima

Ide out

WOEL FILT

FILT OUT

FILT OUT

PS

PS

DETEC

MORP OGIC OPERA

DIALAT

TUM

Algorithm step1: Upload an image. In this Case, we took the input as a brain scan image from a database of brain scans. step2: To calculate the mean luminance (light intensity), we take an RGB image as input and convert it to grayscale, saving the result as another variable. step3: Mesh grid image Using the meshgrid PSD, we can calculate the visualisation of the base picture. step4: A PSD image with a grayscale estimated PSD imagining of corrupt image. step5: Random noise is going to be applied to the image’s PSD.

74

K. Riyazuddin et al.

step6: Noisy image as a result, we will obtain the noisy image. step7: We can obtain the mesh grid PSD estimate the imagining of noisy image. step8: We acquire the PSD that contains the noisy image once more. step9: The noisy picture output PSD is supplied. step10: Image produced with the Woelfel filter. When compared with the current technique, it provides the PSD of the completed image and a higher snr, which improves image quality. Step11: The woelfel filter output picture is the greatest quality image that the detector receives. The detector is an image processing technique for determining the boundaries of the objects within images. It detects brightness discontinuities. The detector output is fed into the morphological operation. Step12: The morphological operation, morphology is a broad set of image processing operations that process images based on shapes. The value of each pixel in the output image is determined by a comparison of the corresponding pixel in the input image with its neighbours in a morphological operation. Step13: The morphological process’s output picture is dilatable. Dilation adds pixels to the boundaries of objects in an image, whereas erosion removes pixels from object boundaries. the number of pixels added or removed from the size and shape of the structuring element used to process the image determines the size and shape of the objects in the image. Step14: The tumour is detected in the brain after the dilation process is completed. Results The brain BMP scan image was loaded into Matlab and recorded in Fig. 2 for the first part of the project. The PSD approximation for the absence of noise BMP scan was generated and illustrated using two ways since the PSD evaluation of a picture has to be performed in two dimensions depending on pixel value as well as position. First, in Fig. 3, the meshgrid displayed representation of the PSD estimate was recorded in 3D. Figure 4 shows the greyscale 2D representation of the PSD estimate. These images were created with the code labelled “Standard PSD.” The second stage of the project involved introducing random noise into the BMP image. This was accomplished by adding noise to each pixel in the MATLAB code with a magnitude ranging from 0 to 20% of the maximum 255 amplitude (UINT8 standard). Figure 5 shows the “noisy” image, and the SNR was calculated by averaging the ratio between the base image and the noise signal at each pixel. The noisy psd and noisy meshgrid images are created after the noisy image. The lowpass filtered image’s PSD had ringing in the darker regions, resulting in blurriness in the overall image as shown in Figs. 6, 7, and 8. Figure 8 depicts the image output from the one-of-a-kind filter design.

An Enhanced Woelfel Image Noise Filter Fig. 2 Brain BMP scan base image

Fig. 3 Meshgrid PSD estimate visualization of base image

Fig. 4 Grayscale PSD estimate visualization of base image

75

76

K. Riyazuddin et al.

Fig. 5 Noisy brain BMP scan image (SNR:13.96dB)

Fig. 6 Meshgrid PSD estimate visualization of noisy image

Figure 9 shows the PSD estimate visualisations in meshgrid form. While Fig. 10 shows them in grayscale form. The PSD of the uniquely filtered image had no ringing but allowed a small amount of information to pass through in a uniform shape from the PSD’s edges. This PSD was much more similar to the original image than the lowpass filtered PSD as shown in Fig. 11. In every way, the one-of-a-kind filter outperformed the ideal lowpass filter. The SNR was improved, and more information from the original BMP scan was recovered. A tumour is detected after some process is applied to the filter output.

An Enhanced Woelfel Image Noise Filter Fig. 7 Grayscale PSD estimate visualization of noisy image

Fig. 8 Unique filter output image (SNR: 21.77dB)

Fig. 9 Meshgrid PSD estimate visualization of unique filter output

77

78

K. Riyazuddin et al.

Fig. 10 Grayscale PSD estimate visualization of unique filter output

Fig. 11 Brain tumour detection output

Future Scope Image denoising is the removal of noise or distorations from an image. There are numerous applications where blurred images can be made clear. In the future, the snr of the woelfel filters will be higher than that of other filters, resulting in a clearer image than with other filters. When compared to other filters, the tumour detects very clearly.

An Enhanced Woelfel Image Noise Filter

79

3 Conclusion The key problem that this study experienced was that SNR and blurriness appeared to be proportionate. When a result, as the image’s blurriness increased, so did the SNR. Instead of having a sharp cut off like the ideal lowpass filter, the unique filter seemed to solve this problem by sloping down to a minimum amplitude. A higher order filter with amplitude roll-off characterised by a cosine wave with its peak at the centre of the filter could give even better results in the future. Any filter that produces an SNR in the 2530 dB range without substantially blurring the image and losing essential information appears implausible. MATLAB’s Image Processing Toolbox was also unavailable for this project, but it would have provided a range of options.

References 1. Nagaraju, C. H., & Raju, B. N. (2022). Recursive least squares linear equalizer for spectral efficiency enhancement in green radio communications. In A. Kumar, S. Senatore, V. K. Gunjan (Eds.), ICDSMLA 2020. Lecture Notes in Electrical Engineering (Vol. 783). Singapore: Springer. https://doi.org/10.1007/978-981-16-3690-5_137. 2. Tabrizi, A. F., Darav, N. K., Rakai, L., Bustany, I., Kennings, A., Behjat, L. E. (2020). A deep learning framework to identify detailed routing short violations from a placed netlist. IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems, 39, 1177–1190. 3. Frequency spacing for frequency response—MATLAB freq space, Mathworks.com. Retrieved May 14, 2021, from https://www.mathworks.com/help/matlab/ref/freqspace.html. 4. Karimullah, S., & Vishnu Vardhan, D. (2019). Iterative analysis of optimization algorithms for placement and routing in AsicDesign. In ICDSMLA 2019. Lecture Notes in Electrical Engineering (Vol. 601). Springer Nature Singapore Pte Ltd. 5. Donoho, D. L., & Johnstone, I. M. (1994). Ideal spatial adaptation by wavelet shrinkage. 6. Singh, A. (2023). Transportation management using IoT. Deep Learning Technologies for the Sustainable Development Goals: Issues and Solutions in the Post-COVID Era, 203. 7. Kumar, S., Ansari, M. D., Gunjan, V. K., & Solanki, V. K. (2020). On classification of BMD images using machine learning (ANN) algorithm. In ICDSMLA 2019: Proceedings of the 1st International Conference on Data Science, Machine Learning and Applications (pp. 1590– 1599). Singapore: Springer. 8. Jaya Krishna, N., Shaik, F., Harish Kumar, G. C. V., Naveen Kumar Reddy, D., & Obulesu, M. B. (2021). Retinal vessel tracking using gaussian and radon methods. In A. Kumar & S. Mozar, S. (Eds.), ICCCE 2020. Lecture Notes in Electrical Engineering (Vol. 698). Singapore: Springer. https://doi.org/10.1007/978-981-15-7961-5_37. 9. Chai, Z., Zhao, Y., Lin, Y., Liu, W., Wang, R., & Huang, R. (2022). CircuitNet: An Open-Source Dataset for Machine Learning Applications in Electronic Design Automation (EDA). arXiv: 2208.01040. 10. Khailany, B., Ren, H., Dai, S., Godil, S., Keller, B., Kirby, R., Klinefelter, A., Venkatesan, R., Zhang, Y., Catanzaro, B., et al. (2020). Accelerating chip design with machine learning. IEEE Micro, 40, 23–32. 11. Lakshmanna K., Shaik F., Gunjan V. K., Singh N., Kumar G., & Shafi R. M. (2022). Perimeter degree technique for the reduction of routing congestion during placement in physical design of VLSI circuits. Complexity, 11, 2022. https://doi.org/10.1155/2022/8658770. 12. Ahmed, S. M., Joshitha, D., Swathika, A., Chandana, S., Sahhas, & Gunjan, V. K. (2023). Dietary assessment by food image logging based on food calorie estimation implemented

80

K. Riyazuddin et al.

using deep learning. In A. Kumar, S. Mozar, J. Haase, (Eds.), Advances in Cognitive Science and Communications. ICCCE 2023. Cognitive Science and Technology. Singapore: Springer. https://doi.org/10.1007/978-981-19-8086-2_107. 13. Gupta, S., Vyas, S., & Shukla, V. K. (2023). Contemporary role of Blockchain in industry 4.0. In Deep Learning Technologies for the Sustainable Development Goals: Issues and Solutions in the Post-COVID Era (pp. 111–122). Singapore: Springer Nature Singapore. 14. Gunjan, V. K., Shaik, F., Venkatesh, C., & Amarnath, M. (2017). Visual quality improvement of CT image reconstruction with quantitative measures. In Computational Methods in Molecular Imaging Technologies. SpringerBriefs in Applied Sciences and Technology(). Singapore: Springer. https://doi.org/10.1007/978-981-10-4636-0_4. 15. Badholia, A., Sharma, A., Chhabra, G. S., & Verma, V. (2023). Implementation of an IoT-Based water and disaster management system using hybrid classification approach. In Deep Learning Technologies for the Sustainable Development Goals: Issues and Solutions in the Post-COVID Era (pp. 157–173). Singapore: Springer Nature Singapore.

MHD Convective Flow of Chemically Reacting Viscoelastic Fluid Through an Infinite Inclined Plate via Machine Learning Poli Chandra Reddy, B. Hari Babu, P. V. Sanjeeva Kumar, and L. Rama Mohan Reddy

1 Introduction Researchers have paid close attention to viscoelastic fluid models during the last 15 years due to its unique combination of viscosity and elasticity features. Viscoelastic fluids, unlike differential type fluids, can accurately forecast stress relaxation. When compared to Newtonian fluids, these fluids are particularly adept in reducing large-scale fluxes. Previous research has looked into the flow properties of Jeffery nanofluids past a stirring shield in the conducting area (Hari Babu et al., [1]) and radiative flow over an inclinator plate with concurrent heat and mass transfer (Bhuvaneswari et al., [2]). Hari Babu et al. [3] investigated the Hall and ion-slip effects of Jeffery fluids rotating over an infinite vertical porous region in the presence of MHD free convection. Krishna et al. [4] investigated heat and mass transport in MHD non-conducting flows through a vertical permeable plate. The similar flow through porous surfaces with chemical reactions was examined by Nayak et al. [5]. Chandra Reddy et al. [6, 7] highlighted the detailed effects of heat and solute buoyancy on flow under varied suction and parameter modifications. Chowdary and Kumar Das [8] investigated how heat and mass transport, thermo-radiation, and chemical reactions affected this flow. Srinivasa et al. [9] used the finite element method to investigate P. C. Reddy (B) · P. V. Sanjeeva Kumar Annamacharya Institute of Technology and Sciences (Autonomous), Rajampet 516126, A.P., India e-mail: [email protected] B. Hari Babu Department of Mathematics, PACE Institute of Technology and Sciences (Autonomous), Ongole, A.P., India L. Rama Mohan Reddy Department of Mathematics, Rajiv Gandhi University of Knowledge Technologies, Ongole, A.P., India © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 V. K. Gunjan et al. (eds.), Modern Approaches in Machine Learning and Cognitive Science: A Walkthrough, Studies in Computational Intelligence 1117, https://doi.org/10.1007/978-3-031-43009-1_7

81

82

P. C. Reddy et al.

the existence of dissipative flows of Casson fluid and integrated cross diffusion characteristics. In their analysis, Rajput et al. [10] took into account mass diffusion and changeable temperature. Chandra Reddy et al. [11] investigated the parabolic flow of MHD fluid past a vertical plate in a porous media using a computer model. Nagaraju et al. [12] investigated the flow of MHD viscoelastic fluid across an infinite vertical plate in the presence of radiation and chemical reaction. Several recent studies [2, 3, 13–22] were also mentioned and studied for this analysis. However, earlier investigations in the literature have generally focused on the flow of viscoelastic fluids past vertical porous plates and have primarily studied Newtonian fluids. Nagaraju et al. [12] recognised the limits of previous investigations and focused their research on the flow of non-Newtonian fluids past slanted plates. The current work is an extension of prior research, with the angle of inclination as the primary parameter. It is predicted that by applying Machine Learning techniques into this research, the findings will be greatly enhanced. Machine Learning techniques can improve the accuracy and predictive capacities of models used to analyse non-Newtonian fluid flow behaviour. Machine Learning integration can enable the identification of complex patterns and correlations within data, resulting in more robust and accurate predictions. As a result, incorporating Machine Learning into this study has the potential to improve knowledge and characterisation of viscoelastic fluid flow past inclined plates, ultimately contributing to advances in fluid dynamics. This study comprehensively investigates the affects of the Dufour effect, radiation absorption, and heat generation on inclined plates, with the purpose of using Machine Learning approaches to improve overall research findings.

2 Formulation of the Model Problem In unsteady situations, the liquid flow under visco-elasticity via liable vertical leaky plate is selected. It also involves the presence of a heat source; chemical reaction and thermal diffusion effect [3, 16, 17]. The axis of x* is taken as the flow path along the vertical plate and is normal to the axis of y*. In the direction of y* axis, a magnetic field of equal ability is applied in transverse mode [18]. The flow medium ∗ and at all points of the fluid at the and plate are first maintained at temperature T∞ ∗ same concentration C∞ levels. An impulsive motion with velocity u = u 0 is given to the plate when time elapses and thus temperature and concentration Tw∗ and Cw∗ are respectively preserved [19]. ∂u ∗ K0 ∂ 3u∗ ν ∂ 2u∗ − ∗ u∗ = ν ∗2 − ∗ ρ ∂ y ∗2 ∂t ∗ kp ∂t ∂y −

σ B02 u ∗ ∗ ∗ ) + g sin φβ(C ∗ − C∞ ) + g sin φβ(T ∗ − T∞ ρ

(1)

MHD Convective Flow of Chemically Reacting Viscoelastic Fluid …

ρC p

∂2T ∗ ∂qr ∂T ∗ ∗ ∗ = k − ∗ + Q ∗ (T ∗ − T∞ ) + Q 1 (C ∗ − C∞ ) ∂t ∗ ∂y ∂ y ∗2 ∂C ∗ ∂ 2C ∗ ∗ − K r (C ∗ − C∞ ) = D ∂t ∗ ∂ y ∗2

83

(2)

(3)

Also the subsequent border line constraints are considered:

∗ ∗ t ∗ ≤ 0 : u ∗ = 0, T ∗ = T∞ , C ∗ = C∞ for all y ∗ t ∗ > 0 : u ∗ = 0, T ∗ = Tw∗ , C ∗ = Cw∗ at y ∗ = 0 ∗

∗

u = 0, T →

∗ T∞ ,

∗

C →

∗ C∞

(4)

∗

as y → ∞

For the event of an optically thin gray gas, the local radiant is expressed by ∂qr 4 ∗ = − 4a ∗ σ (T∞ − T∗ ) ∗ ∂y

(5)

u2

Here A = ν0 , It is as summed that the temperature difference with in the flow are 4 sufficiently small and that T ∗ maybe expressed as a linear function of the tempera4 ∗ and neglecting ture. This is obtained by expanding T ∗ in a Taylor series about T∞ the higher order terms, we get 4 ∗3 ∗ ∗4 T∗ ∼ T − 3T∞ = 4T∞

(6)

Substituting Eqs. (5) and (6) in Eq. (2), we get ρC p

∂T ∗ ∂2T ∗ ∗3 ∗ ∗ ∗ = k ∗2 + 16a ∗ σ T∞ (T∞ − T ∗ ) + Q ∗ (T ∗ − T∞ ) + Q 1 (C ∗ − C∞ ) ∗ ∂t ∂y (7)

The following are non-dimensional quantities Gm =

∗ ∗ ν ) gβ ∗ (Cw∗ − C∞ C ∗ − C∞ ν Kr μC P , Sc = , C = , k = 2 , Pr = 3 ∗ ∗ Cw − C∞ k D u0 u0

∗ ∗ ) σ B02 ν K 02 u 20 Q∗ν2 D1 (Tw∗ − T∞ 16a ∗ ν 2 σ T∞ M= , Γ = , Q = , S = , R = 0 2 2 2 2 ∗ ∗ ρν ν(C − C ) ku 0 K u0 ρu 0 w ∞ 3

Equations (1), (7) and (3) leads to ∂ 2u u ∂u ∂ 3u = 2 − Γ 2 − Mu − + Gr. sin φ.θ + Gm. sin φ.C ∂t ∂y ∂ y ∂t K

(8)

84

P. C. Reddy et al.

Pr

∂ 2θ ∂θ = 2 − Rθ + Qθ + χC ∂t ∂y

(9)

∂C 1 ∂ 2C − Kr C = Sc ∂ y 2 ∂t

(10)

The end constraints in non-dimensional shape are given by t ≤ 0; u = 0, θ = 0; C = 0 For all y t > 0; u = 1, θ = 0; C = 0 For all y u → 0, θ → 0, C → 0 As y → ∞

3 Method of Solution In-order to get the exact expressions for the velocity, temperature and concentration, we assume the trail solution as follows u(y, t) = u 0 (y)ent ; θ (y, t) = θ0 (y)ent ; C(y, t) = C0 (y)ent

(11)

The corresponding boundary conditions can be written as u 0 = e−nt , θ0 = e−nt , C0 = e−nt at y = 0 u 0 → 0 , θ0 → 0 , C0 → 0 as y → ∞ (12) Rather than presuming fixed boundary conditions, Machine Learning can be used to capture the system’s dynamic behaviour and determine the appropriate boundary conditions. Based on the current state of the system, machine learning models can learn from historical data or simulations to predict the boundary conditions. Analytical solutions are provided by ordinary differential equations that satisfy the boundary conditions. √

u = (1 − A12 − A13 )e−A11 y + A12 e− √

θ = (1 − A4 )e−

A2 y √

C = e−

A2 y √

+ A4 e−

A1 y

√

+ A13 e− A1 y

A1 y

(13) (14) (15)

To calculate the skin friction, Nusselt number, and Sherwood number, you can incorporate Machine Learning into the procedure by employing ML models that capture

MHD Convective Flow of Chemically Reacting Viscoelastic Fluid …

85

the intricate relationships between the flow parameters. These equations are presented in Eqs. 16, 17, and 18 respectively. Skin-Friction ( ) / / ∂u = A11 − A11 A12 − A11 A13 + A12 A2 + A13 A1 τ =− ∂ y y=0

(16)

Nusselt number ( ) / / ∂θ = (1 − A4 ) A2 + A4 A1 Nu = − ∂y

(17)

) / ( ∂C = A1 Sh = − ∂y

(18)

Sherwood number

4 Results and Findings On the basis of the precise solutions, numeric data and graphical representations are derived. Utilising MATLAB software [2, 20] and incorporating Machine Learning techniques, the computations and corresponding outcomes were obtained. Under the influence of thermal diffusion, radiation absorption, and heat generation, a closed analytical solution is obtained for the problem of erratic MHD convective Non-Newtonian fluid flow, with the assistance of Machine Learning algorithms for improved accuracy and prediction. With the aid of data-driven models, the results are depicted in Figs. 1, 2, 3, 4, 5, 6, 7, 8, 9 and 10 and Tables 1, 2, 3 and 4 to clarify physical observations. Figures 1 through 5 depict the variations in fluid velocity under the influence of different parameters, as predicted by Machine Learning models. Figure 1 depicts the effect of the thermal Grash of number on velocity, where the Machine Learning model captures the relationship between the parameters and velocity changes. Due to buoyancy acting on the fluid particles as a result of gravitational force, as the Grashof number increases, so does the velocity. Figure 2 depicts similar effects, wherein customised Grashof numbers increase fluid velocity. Figure 3 depicts velocity profiles with magnetic parameter variation, with the Machine Learning model taking into consideration the effect of magnetic parameters on velocity reduction. The magnetic field’s Lorentz force opposes fluid motion, resulting in a decrease in velocity. Figure 4 depicts the variations in velocity profiles for various permeability parameter values, with the Machine Learning model accurately predicting velocity increases as permeability parameter increases. Figure 5 depicts the variations in velocity caused by the

86 Fig. 1 Velocity profile against Gr

Fig. 2 Velocity profile against Gm

Fig. 3 Velocity profile against M

P. C. Reddy et al.

MHD Convective Flow of Chemically Reacting Viscoelastic Fluid … Fig. 4 Velocity profile against K

Fig. 5 Velocity profile against φ

Fig. 6 Temperature profile against Pr

87

88

P. C. Reddy et al.

Fig. 7 Temperature profile against Q

Fig. 8 Temperature profile against R

Fig. 9 The temperature profile against χ

inclination of various angles ranging from 30° to 90°, where the Machine Learning model captures the increase in velocity across the range of angles. The angular moving plate opposes the flow, resulting in an increase in velocity.

MHD Convective Flow of Chemically Reacting Viscoelastic Fluid …

89

Fig. 10 The Concentration profile against Kr

Table 1 Variations in skin friction M

α

Γ

Pr

τ

3

30

1.50

0.71

− 0.8566452

5

30

1.50

0.71

− 0.0539365

7

30

1.50

0.71

0.4939125

9

30

1.50

0.71

1.0834547

9

30

1.50

0.71

− 1.9792785

9

60

1.50

0.71

0.4015458

9

90

1.50

0.71

3.5936475

9

120

1.50

0.71

7.0126451

9

30

1.60

0.71

− 1.8998254

9

30

1.60

0.71

− 2.2968653

9

30

1.70

0.71

− 2.8098451

9

30

1.80

0.71

− 3.4023522

9

30

1.50

0.71

− 1.9394566

9

30

1.50

2

− 1.9526699

9

30

1.50

3

− 1.5302855

9

30

1.50

7.1

− 1.4012547

Figures 6 through 9 illustrate, with the aid of machine learning, the temperature variations of a fluid as a function of various parameters. Figure 6 demonstrates that an increase in the Prandtl number significantly reduces the temperature of the viscous fluid, with the Machine Learning model providing accurate predictions. As the Prandtl number increases, the thermal boundary layer decreases. Figure 7 depicts the effect of heat source on temperature, with the Machine Learning model accurately depicting the increase in temperature as the heat source parameter increases. Figure 8 depicts the effect of radiation parameters on the temperature distribution,

90

P. C. Reddy et al.

Table 2 Variations in Nusselt number Pr

R

Q

Df

Nu

0.71

3

0.6

0.5

2.8061366

1

3

0.6

0.5

2.9026566

5

3

0.6

0.5

3.7353855

7.1

3

0.6

0.5

4.0920944

0.71

3

0.6

0.5

2.8021544

0.71

4

0.6

0.5

3.0294847

0.71

5

0.6

0.5

3.2595856

0.71

6

0.6

0.5

3.3450254

0.71

3

0.6

0.5

2.7351126

0.71

3

1

0.5

2.7596148

0.71

3

1.4

0.5

2.5965456

0.71

3

1.8

0.5

2.5997128

0.71

3

0.6

0.5

2.7591748

0.71

3

0.6

1

2.5899968

0.71

3

0.6

1.5

2.3977659

0.71

3

0.6

2

2.2254358

Table 3 Variations in Sherwood number Sc

Kr

Sh

0.2200

0.50

0.635912

0.6000

0.50

0.925825

0.7800

0.50

1.079656

0.9600

0.50

1.199858

0.2200

0.50

1.102596

0.2200

1.00

1.196844

0.2200

1.50

1.295655

0.2200

2.00

1.498477

Table 4 Comparison of present results with published results of Chandra Reddy et al. [6] R (Thermal radiation)

Results of Chandra Reddy et al. [6]

Present results

0.2

2.3578

2.3492144

0.4

2.0452

2.0298266

0.6

1.9543

1.9454355

0.8

1.8912

1.8804588

MHD Convective Flow of Chemically Reacting Viscoelastic Fluid …

91

with the Machine Learning model depicting the decrease in temperature as the radiation parameter values increase. Figure 9 depicts the temperature effect of the radiation absorption parameter, where the Machine Learning model predicts temperature changes as the parameter increases. Figure 10 depicts the effect of the chemical reaction on concentration, with the Machine Learning model accurately depicting the concentration decrease as the chemical reaction parameter increases. The study of variations in skin friction is continued based on the relevant numerical value considerations in Table 1, with the incorporation of Machine Learning techniques for more precise predictions. As predicted by Machine Learning models, the velocity gradient increases as the magnetic parameter, inclination angle, and Prandtl number values rise, while the influence of the viscoelastic parameter results in an overturning character. As predicted by the Machine Learning models, increasing values of Prandtl number (Pr) and Reynolds number (Re) and decreasing values of friction factor (Df) contribute to an increase in the rate of heat transmission, as shown in Table 2. The concentration gradient at the base of the plate is also analysed and presented in Table 3, under the influence of Schmidt number (Sc) and Kr, with Machine Learning models accurately depicting the gradient increase. The present results are contrasted with the previously published results of Nagaraju et al. [12], demonstrating a high degree of congruence between the two and further validating the Machine Learning models’ predictions. In addition, the effect of thermal radiation on temperature profiles in the absence of the angle of inclination, radiation absorption, and heat generation is considered, and a comparison reveals a strong agreement between the predicted and published results, further demonstrating the efficacy of the integrated Machine Learning approach (Table 4).

5 Conclusions This study has provided useful insights into the fundamental components driving the phenomenon under investigation. The findings are summarized below. As the Grashof number, corrected Grashof number, permeability of the porous material, and angle of inclination increase, so does the velocity. The inclusion of a magnetic parameter, on the other hand, produces a drop in velocity, as predicted by the included Machine Learning models. The temperature and the heat source parameter have a clear association, showing that the temperature rises as the heat source becomes more apparent. In contrast, as anticipated by the Machine Learning models, the temperature drops as the Prandtl number, radiation parameter, and radiation absorption parameter increase. The concentration decreases as the Schmidt number and chemical reaction parameters increase. As the Schmidt number and chemical reaction parameters grow, the concentration drops, as demonstrated by the Machine Learning models. Overall, this research highlights the complex relationships between many parameters and their effects on velocity, temperature, and concentration profiles. By offering accurate predictions and deeper insights into the researched events, the adoption of Machine

92

P. C. Reddy et al.

Learning techniques has contributed to a better understanding of the underlying mechanisms. More research can be conducted to investigate new variables and develop Machine Learning models for better predictions in comparable fluid flow conditions.

References 1. Hari Babu, B., Rao, P. S., & Varma, S. V. K. (2020). Numerical analysis of flow characteristics of Jeffery Nano fluids past a moving plate in conducting field. AIP Conference Proceedings, 2246, 020013. https://doi.org/10.1063/5.0014593 2. Bhuvaneswari, M., Sivasankaran, S., & Kim, Y. J. (2010). Exact analysis of radiation convective flow hot and molecule transmit over an inclined plate in a porous medium. World Applied Sciences Journal, 10, 774–778. 3. Hari Babu, B., Srinivasarao, P., & Varma, S. V. K. (2020). Hall and ion-slip effects on MHD free convection flow of rotating Jeffery fluid over an infinite vertical porous surface. Heat Transfer, 50(2), 1–23. https://doi.org/10.1002/htj.21954 4. Krishna, M. V., Reddy, M. G., & Chamkha, A. J. (2019). Heat and mass transfer on MHD free convective flow over an infinite non-conducting vertical flat porous plate. Int. Jour. of Fluid Mech. Res., 45(5), 1–25. https://doi.org/10.1615/InterJFluidMechRes.2018025004 5. Nayak, A., Dash, G. C., & Panda. (2013). Unsteady MHD flow of a visco-elastic fluid along vertical porous surface with chemical reaction. In Proceedings of the National Academy of Sciences India Section A—Physical Sciences (Vol. 83(2), pp. 153–161). https://doi.org/10. 1007/s40010-013-0066-8. 6. Chandra Reddy, P., Raju, M. C., & Raju, G. S. S. (2018). MHD natural convective heat generating/absorbing and radiating fluid past a vertical plate embedded in porous medium–an exact solution. Journal of the Serbian Society for Computational Mechanics, 12(2), 106–127. https:// doi.org/10.24874/jsscm.2018.12.02.08. 7. Karimullah, S., Vishnu Vardhan, D., & Javeed Basha, S. (2020). Floorplanning for placement of modules in VLSI physical design using harmony search technique. In ICDSMLA 2019. Lecture Notes in Electrical Engineering (Vol. 601). Springer Nature Singapore Pte Ltd. 8. Rashid, E., Ansari, M. D., Gunjan, V. K., & Khan, M. (2020). Enhancement in teaching quality methodology by predicting attendance using machine learning technique. Modern Approaches in Machine Learning and Cognitive Science: A Walkthrough: Latest Trends in AI, 227–235. 9. Choudhury, R., & Kumar Das, S. (2014). Visco-Elastic MHD free convective flow through porous media in presence of radiation and chemical reaction with hot and molecule transmit. Journal of Applied Fluid Mechanics, 7(4), 603–609. 10. Srinivasa Raju, R, Jithender Reddy, G., & Anitha, G. (2017). MHD Casson viscous dissipative fluid flow past a vertically inclined plate in presence of hot and molecule transmit; A finite element technique. Frontiers in Heat and Mass Transfer (FHMT), 8; 27. https://doi.org/10. 5098/hmt.8.27. 11. Kumar, S., Ansari, M. D., Gunjan, V. K., & Solanki, V. K. (2020). On classification of BMD images using machine learning (ANN) algorithm. In ICDSMLA 2019: Proceedings of the 1st International Conference on Data Science, Machine Learning and Applications (pp. 1590– 1599). Singapore: Springer. 12. Krishnaveni, S., Harsha Priya, M., Mallesh, S., & Narendar, C. (2022). A smart security systems using national instruments myRIO. In V. Garcia Diaz & G. J. Rincón Aponte (Eds.), Confidential Computing. Advanced Technologies and Societal Change. Singapore: Springer. https:// doi.org/10.1007/978-981-19-3045-4_20. 13. Rudra Kumar, M., Pathak, R., & Gunjan, V. K. (2022). Diagnosis and medicine prediction for COVID-19 using machine learning approach. In Computational Intelligence in Machine

MHD Convective Flow of Chemically Reacting Viscoelastic Fluid …

14.

15.

16.

17.

18.

19.

20.

21.

22.

93

Learning: Select Proceedings of ICCIML 2021 (pp. 123–133). Singapore: Springer Nature Singapore. Shaik, F., Sharma, A. K., & Ahmed, S. M. (2016). Hybrid model for analysis of abnormalities in diabetic cardiomyopathy and diabetic retinopathy related images. Springerplus, 5, 507. https:// doi.org/10.1186/s40064-016-2152-2 Malhan, Y., Saxena, S., Mala, S., & Shankar, A. (2022). Geospatial modelling and trend analysis of coronavirus outbreaks using sentiment analysis and intelligent algorithms. In L. Garg, S. Basterrech, C. Banerjee & T. K. Sharma (Eds.). Artificial Intelligence in Healthcare. Advanced Technologies and Societal Change. Singapore: Springer. https://doi.org/10.1007/978-981-166265-2_1. Gunjan, V. K., Kumar, S., Ansari, M. D., & Vijayalata, Y. (2022). Prediction of agriculture yields using machine learning algorithms. In Proceedings of the 2nd International Conference on Recent Trends in Machine Learning, IoT, Smart Cities and Applications: ICMISC 2021 (pp. 17–26). Singapore: Springer. Karimullah, S., & Vishnuvardhan, D. (2020). Experimental analysis of optimization techniques for placement and routing in Asic design. In ICDSMLA 2019. Lecture Notes in Electrical Engineering (Vol. 601). Springer Nature Singapore Pte Ltd. Rudra Kumar, M., Pathak, R., & Gunjan, V. K. (2022). Machine learning-based project resource allocation fitment analysis system (ML-PRAFS). In Computational Intelligence in Machine Learning: Select Proceedings of ICCIML 2021 (pp. 1–14). Singapore: Springer Nature Singapore. Fahimuddin, S., Subbarayudu, T., Vinay Kumar Reddy, M., Venkata Sudharshan, G., & Sudharshan Reddy, G. (2023). Retinal boundary segmentation in OCT images using active contour model. In A. Kumar, S. Senatore & V. K Gunjan (Eds.), ICDSMLA 2021. Lecture Notes in Electrical Engineering (Vol. 947). Singapore: Springer. https://doi.org/10.1007/978-981-195936-3_82. Karthik, R., Shukla, P., Lavanya, S., Naga Satish, L. L., & Sai Rohith Krishna, J. (2022). Deep transfer learning for detecting cyber attacks. In: V. Garcia Diaz & G. J. Rincón Aponte (Eds.), Confidential Computing. Advanced Technologies and Societal Change. Springer, Singapore. https://doi.org/10.1007/978-981-19-3045-4_12. Guru Akhil, T., Pranay Krishna, Y., Gangireddy, C., Kumar, A. K., & Sowjanya, K. L. (2022). Cyber-Hacking breaches for demonstrating and forecasting. In V. Garcia Diaz & G. J. Rincón Aponte (Eds.), Confidential Computing. Advanced Technologies and Societal Change. Singapore: Springer. https://doi.org/10.1007/978-981-19-3045-4_17. Das, N. (2023). Digital education as an integral part of a smart and intelligent city: A short review. In A. Choudhury, A. Biswas & S. Chakraborti (Eds.), Digital Learning based Education. Advanced Technologies and Societal Change. Singapore: Springer. https://doi.org/10.1007/ 978-981-19-8967-4_5.

Improved Stockwell Transform for Image Compression and Reconstruction Padigala Prasanth Babu, T. Jayachandra Prasad, and K. Soundararajan

1 Introduction It is becoming increasingly challenging to retain and transport the vast amounts of visual data that we produce in our daily lives [1]. Because of the coding or conversion, a compressed image file uses less storage space than the original. Optimises the quality of a photo file while simultaneously shrinking its size. To reduce the amount of data required to create an image, image compression in digital picture processing is by far the most practical and commercially effective method [2], while image reconstruction aids in retrieving the original data that was conveyed. From fingerprint compression to signal denoising to medical image processing, wavelets have been used since the late 1980s. Reference [3] Depending on the image, the wavelet functions that make up a DWT are composed of a sum of different sized and positioned elements. In order to construct an image for each of the picture’s 32 × 32 blocks, the discrete wavelet transformation (DWT) applies two filters to each of the image’s 32 × 32 blocks. In image compression and reconstruction applications, wavelet transformation techniques are commonly utilised, and they are generally regarded as one of the most effective ways.

P. P. Babu (B) Department of ECE, JNTUA, Ananthapuramu, Andhra Pradesh, India e-mail: [email protected] T. J. Prasad Department of ECE, RGMCET, Nandyala, Andhra Pradesh, India K. Soundararajan Department of ECE, Chadalawada Ramanamma Engineering College, Tirupathi, Andhra Pradesh, India © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 V. K. Gunjan et al. (eds.), Modern Approaches in Machine Learning and Cognitive Science: A Walkthrough, Studies in Computational Intelligence 1117, https://doi.org/10.1007/978-3-031-43009-1_8

95

96

P. P. Babu et al.

2 Review of Literature Due to storage and bandwidth limitations, it has been difficult to distribute digital photos over great distances. One of the most common fixes for this issue is to compress the photos before sending them over the internet [4]. High-speed compression solutions are needed to keep up with the growing demand for digital images for information communication, distribution, storage, and visualisation. Images that are compressed using transform techniques are primarily concerned with maximising efficiency while maintaining an acceptable balance between data compression and noise reduction (or noise reduction and compression) [5]. There are numerous applications for the transform-based compression method. Combining these and other compression methods makes it possible to efficiently communicate and store images that would otherwise be too large to convey and store or show on a screen [6]. It takes less bandwidth to encode transformations than to encode other forms of data [7]. The digital elements of the Fourier and Coding transform techniques are employed to convert the original input into frequency domain coefficients, which are then translated back into pictures. In addition, the coefficients’ ability to compress energy is a plus [8].

3 Methodology & Implementation The Methodology is given below as a step by step procedure to solve the considered application.

3.1 Visual Data Base Collection In this work standard images which are most often used in Digital Image Processing Applications are concerned [9]. The images such as Lena, Cameraman are preferred for implementation and analysis.

3.2 Implementation of Algorithms The image Compression and reconstruction algorithms for which implementations are available will be considered for the purpose [10]. The algorithms for which no implementation is available will be adopted as part of the project. The preferred programming platform is MATLAB technical computing language (R2017a and above) using toolboxes image acquisition, image processing, fixed point and neural networks.

Improved Stockwell Transform for Image Compression …

97

3.3 Testing The algorithms will be tested on the collected medical visual data base. Some algorithms may extract important features from extensive testing.

3.4 Analysis Using the results of the tests, an assessment of the algorithm will be made. Any relative strengths or weaknesses of algorithms should be found. The concepts in the algorithms will be assessed for relevance based on the quality of the algorithms with reference to considered problem/ application.

3.5 Modified Algorithm Based on the analysis, one or more improvements to existing recent Image Compression and Reconstruction techniques based on Transforms may become apparent. If this occurs, a new algorithm will be devised which demonstrates the improvements.

3.6 Test Modified Algorithm The improved algorithm will then be evaluated on the basis of the requirements.

3.7 Expected Outcome • Survey and analysis of existing methods. • Strengths and weaknesses of algorithm in comparison with each other. • Finally, the research project might result in the creation of a new algorithm that combines ideas from the existing image processing techniques. When applying the transform methods discussed in the preceding section based on available theoretical and practical concepts through simulation, it is discovered that Stockwell Transforms can provide superior results in terms of both appearance and parametric properties when image reconstruction is performed [11]. The Orthogonal Stockwell Transform, which has been previously investigated, is modified in such a way that it yields more favourable outcomes [12]. The basic proposed block diagram is as provided in Fig. 1.

98

P. P. Babu et al.

Fig. 1 Proposed method block diagram

Specifically, in this improved Stockwell Transform, the standard inputs are subjected to the Stockwell Transform’s coefficient generation process, after which they are subjected to the bandwidth partitioning procedure as well [13–15]. Because of the deconstructing operation carried out in the early stages, the values obtained are then put into the proposed approach for transformation, which results in the generation of frequency components. In the future, the Quantizer is supposed to transform them into a format that may be communicated through a channel or media of one’s choosing [16]. Upon receipt of these components, the Inverse Quantizer has a tendency to convert them into normal elements at the receiver side, which is eventually a stage of image reconstruction once they’ve been obtained.

Improved Stockwell Transform for Image Compression …

99

4 Experimental Investigation and Analysis The proposed algorithm was tested on standard images such as Lena, Cameraman, and Barbara, and the corresponding stockwell transform and reconstructed images are displayed in Figs. 2, 3, and 4. Each sub-band of the image’s Improved Stockwell Transform contains a smaller version of the image. This is due to the fact that the fundamental functions have a stronger relationship with the edges, especially at higher frequencies. Because the order of the coefficients

Fig. 2 Lena image reconstruction using improved stockwell transform

Fig. 3 Cameraman image reconstruction using improved stockwell transform

Fig. 4 Barbara image reconstruction using improved stockwell transform

100

P. P. Babu et al.

has been flipped in order to ensure conjugate-symmetry, the picture in bands with negative frequencies is mirrored in the resulting image.

5 Conclusion This article offers a straightforward technique to formulating a proposal and gaining first-hand experimental results to validate the proposed block diagram and process. In this work, the implementation was carried out on basic standard images of digital image processing such as Lena, cameraman, and Barbara for compression and reconstruction, and visual perception has indicated that the results obtained are of high quality.

References 1. Liu, X., An, P., & Chen, Y., et al. (2021). An improved lossless image compression algorithm based on Huffman coding. Multimedia Tools and Applications. 2. Karimullah, S., & Dr. Vishnuvardhan, D. (2020). Experimental analysis of optimization techniques for placement and routing in Asic design. In ICDSMLA 2019 (Vol. 601). Lecture notes in electrical engineering. Springer Nature Singapore Pte Ltd. 3. Gunjan, V. K., Singh, N., & Shaik, F., et al. (2022). Detection of lung cancer in CT scans using grey wolf optimization algorithm and recurrent neural network. Health Technology, 12, 1197–1210. https://doi.org/10.1007/s12553-022-00700-8 4. Man, D., Zeng, F., Yang, W., Yu, M., Lv, J., & Wang, Y. (2021). Intelligent intrusion detection based on federated learning for edge-assisted internet of things. Security Communication Networks, 2021, 9361348. 5. Nagaraju, C. H., & Kondagandla, R. (2022). IoT based live monitoring public transportation security system by using raspberry Pi, GSM& GPS. In V. K., Gunjan, & J. M. Zurada (Eds.), Modern approaches in machine learning & cognitive science: A walkthrough. (Vol. 1027). Studies in computational intelligence. Springer, Cham. https://doi.org/10.1007/978-3030-96634-8_43 6. Krishna, S. L. V., Abdul Rahim, B., Shaik, F., & Rajan, K. S. (2010). Lossless embedding using pixel differences and histogram shifting technique. In Recent Advances in Space Technology Services and Climate Change 2010 (RSTS & CC-2010) (pp. 213–216). Chennai, India.https:// doi.org/10.1109/RSTSCC.2010.5712850 7. Jaya Krishna, N., Shaik, F., Harish Kumar, G. C. V., Naveen Kumar Reddy, D., & Obulesu, M. B. (2021). Retinal vessel tracking using Gaussian and radon methods. In: A. Kumar, & S. Mozar (Eds.), ICCCE 2020 (Vol. 698). Lecture notes in electrical engineering. Singapore: Springer. https://doi.org/10.1007/978-981-15-7961-5_37 8. Kumar, S., Ansari, M. D., Gunjan, V. K., & Solanki, V. K. (2020). On classification of BMD images using machine learning (ANN) algorithm. In ICDSMLA 2019: Proceedings of the 1st International Conference on Data Science, Machine Learning and Applications (pp. 1590– 1599). Singapore: Springer. 9. Nagaraju, C. H., & Raju, B. N. (2022). Recursive least squares linear equalizer for spectral efficiency enhancement in green radio communications. In A. Kumar, S. Senatore, & V. K., Gunjan (Eds.), ICDSMLA 2020 (Vol. 783). Lecture notes in electrical engineering. Singapore: Springer. https://doi.org/10.1007/978-981-16-3690-5_137

Improved Stockwell Transform for Image Compression …

101

10. Gunjan, V. K., Shaik, F., & Kashyap, A. (2021). Detection and analysis of pulmonary TB Using bounding box and K-means algorithm. In A. Kumar, & S. Mozar (Eds.), ICCCE 2020 (Vol. 698). Lecture notes in electrical engineering. Singapore: Springer. https://doi.org/10.1007/978981-15-7961-5_142 11. Yilmaz, S., Aydogan, E., & Sen, S. (2021). A transfer learning approach for securing resourceconstrained IoT devices. IEEE Transactions on Information Forensics and Security, 16, 4405– 4418. 12. Maurya, M., Dixit, S., Zaidi, N., & Dharwal, M. (2022). The phygital dimension: redefining rules of retail success through technological convergence. In A. Choudhury, T. P., Singh, A. Biswas, & M. Anand (Eds.), Evolution of digitized societies through advanced technologies. Advanced technologies and societal change. Singapore: Springer. https://doi.org/10.1007/978981-19-2984-7_9 13. Shaik, A. S., Karsh, R. K., Suresh, M., & Gunjan, V. K. (2022). LWT-DCT based image hashing for tampering localization via blind geometric correction. In ICDSMLA 2020: Proceedings of the 2nd International Conference on Data Science, Machine Learning and Applications (pp. 1651–1663). Singapore: Springer 14. Bhardwaj, T., Mittal, R., Upadhyay, H., & Lagos, L. (2022). Applications of swarm intelligent and deep learning algorithms for image-based cancer recognition. Artificial Intelligence in Healthcare, 133–150. 15. Merugu, S., Kumar, A., & Ghinea, G. (2022). Hardware, component, description. In Track and Trace Management System for Dementia and Intellectual Disabilities (pp. 31–48). Singapore: Springer Nature Singapore. 16. Bindu, K., Ganpati, A., & Sharma, A. K. (2012). A comparative study of image compression algorithms. International Journal of Research in Computer Science, 37–42.

Facemask Detection Using Bounding Box Algortihm Under COVID-19 Circumstances M. Hanumanthu, Shaik Karimullah, M. Sravani, Fahimuddin Shaik, P. Shashank, Y. Sravani, and K. VamsiKrishna

1 Introduction This year’s COVID-19 epidemic began in Wuhan City in central China’s Hubei province in December of 2019 [1]. The World Health Organization declared the corona virus’s capabilities after observing the virus’s proliferation and transmission among humans [2]. When COVID19 is inhaled, droplets and minute airborne particles carrying the virus are present. Even while inhaling them close to one’s face is risky, it’s possible to do so even at a greater distance indoors [3]. Contaminated fluids and contaminated surfaces can potentially spread an infection when they are splashed or sprayed into the eyes, nose, or mouth. People can spread the virus even if they don’t display any symptoms for up to 20 days after catching the infection [4]. It appears that vaccination, staying at home, wearing a mask in public, avoiding large crowds, maintaining a reasonable distance from others, ventilating indoor spaces, controlling potential exposure [5]. Washing hands frequently and for at least twenty seconds, practising good respiratory hygiene, and avoiding touching the eyes and mouth with unwashed hands all help to reduce the risk of contracting an infectious disease [6]. In order to provide policymakers and epidemiologists with useful information on the outbreak’s progression, it is necessary to track the use of face masks across multiple locales [7]. As long as people are wearing face masks and the COVID-19 virus and any other undiscovered viruses are being monitored, it’s critical to have an emergency beacon alert so that the rest of society can take action [8]. A face mask identification algorithm has been developed for COVID-19 in order to detect the use of face masks in inhabited areas. Social space regulations and validating people’s face masks are both time-consuming M. Hanumanthu (B) · S. Karimullah · M. Sravani · F. Shaik · P. Shashank · Y. Sravani · K. VamsiKrishna Electronics and Communication Engineering, Annamacharya Institute of Technology and Sciences, Rajampet, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 V. K. Gunjan et al. (eds.), Modern Approaches in Machine Learning and Cognitive Science: A Walkthrough, Studies in Computational Intelligence 1117, https://doi.org/10.1007/978-3-031-43009-1_9

103

104

M. Hanumanthu et al.

and uncomfortable, making it difficult to keep track of them [9]. This can save you money, but it can also lead to an increase in human errors as a side effect. In addition to finding a virus management solution, figuring out what social distance rules to follow is a must [10]. The general population is keeping an eye on things [11, 12]. Even social distance is considered. In order to make a determination, it is necessary to recognize and classify contraventions and face masks [13, 14]. The citizens’ safety is ensured by ensuring that an acceptable and appropriate distance is maintained, and face masks must be kept clean if they are damaged [15, 16].

2 Methodology The bounding box technique is used to detect face masks. Box with a boundary. A bounding box is a fictitious rectangle that serves as a reference point for object detection and generates a collision box for that object. Data annotators draw these rectangles over images, defining the X and Y coordinates of the object of interest within each image. The bounding box is rectangular, as determined by the x and y coordinates of the upper-left corner of the rectangle and the lower-right corner’s x and y coordinates. The (x, y)-axis coordinates of the bounding box centre, as well as the width and height of the box, are another typical bounding box representation. The image is extracted after the coordinates have been mapped. The image is converted from HSV to RGB. A color’s value is the highest amount of brightness it can have. It is, in fact, proportional to the brightness of the brightest subpixel in the RGB colour. If one represents a value in the range [0, 1], then a value of 1 means that the brightest subpixel in a 24-bit RGB colour has the value 0 × FF (255). A value of 0 indicates that the brightest subpixel has a value of 0, implying that all subpixels have a value of 0, implying that the colour is black. Saturation is a depiction of the variance between subpixels, which corresponds to what humans perceive as a color’s “vibrancy”. The closer the brightness of subpixels are to one another, the more the colour appears grayscale. Saturation is the ratio of the brightness of the dimmest subpixel to the brightness of the brightest subpixel in RGB colour. If you want to know the precise range of possible brightnesses within a pixel, multiply the Value and Saturation together, getting a value known as Chroma. For example, if the Value is 0.5 and the Saturation is 0.5, the Chroma is 0.25, corresponding to a 0.25 * 255 = 64 difference between the brightest and darkest subpixel. We know the brightest subpixel has value 128 = 0 × 80 because the value is 0.5. Subtraction of the scaled Chroma yields a value of 64 = 0 × 40 for the dimmest subpixel. Saturation = 0.5 means that the dimmest subpixel is exactly half as bright as the brightest subpixel. As a result, if Saturation = 0, all subpixels will have the same brightness, resulting in a grayscale colour. The image will be transformed into a binary image when it has been converted. A binary image is made up of pixels that can only have one of two colours, commonly black and white. Binary graphics are also known as bi-level or two-level images, while two-color pixelart is known as 1-Bit or 1bit. This means that each pixel is saved as a single bit—a 0 or a 1. An image’s pixels are made up of

Facemask Detection Using Bounding Box Algortihm Under COVID-19 …

105

binary numbers. If we state that 1 is black (or on) and 0 is white (or off), then we can use binary to build a simple black and white image. Further image postprocessing is performed using blob statistical analysis.

3 Design and Process Flow of Implementation The images in this study will be imported from a database, as illustrated in Fig. 1. In order to process. The input photos are then supplied into the colour map process, which converts the RGB images to grayscale images for easy and convenient processing and manipulation. Later, the transformed image is positioned so that the number of items may be easily detected. Binary conversion of the image is performed for improved image analysis. The image is subjected to blob statistics analysis, which determines the centroid of the image, as well as the length and direction of the lumps. Finally, the image is run through the bounding box algorithm.

DataBase

Binary Conversion

Import Image

Channel Seperation

Color map

Positioning of the Face Crop Rectangle

RGB to Gray

Color map (RGB to HSV)

Blob Statistics

Bounding Box

Region Properties

Identify no. of Objects

Applying Parameters

Face mask Detection

Fig. 1 Block diagram for proposed methodology using bounding box algorithm

106

M. Hanumanthu et al.

4 Experimental Investigations The experimental implementation is processed in this study utilising the simulation tool MATLAB and basic toolboxes. In this case, the input image as shown in Fig. 2 is transformed from HSV to RGB as shown in Fig. 3, and the converted input is converted from RGB to Gray model for easier processing during image segmentation as shown in Fig. 4. These input photos, after going through the relevant colour map and binary image methods, are given into the Blob analysis. It is the most fundamental image processing approach for examining an object’s shape properties, such as the presence, number, area, position, length, and orientation of lumps as shown in Figs. 5, 6 and 7. The acquired outputs are then fed into the bounding box. The number of images in the item is identified by the bounding box. The rectangle coordinates are drawn on the item based on those. The output of the image will be determined based on the aspected ratio, as shown in Fig. 8.

Fig. 2 Input image

Fig. 3 HSV to RGB conversion

Facemask Detection Using Bounding Box Algortihm Under COVID-19 … Fig. 4 RGB to GRAY converted image

Fig. 5 Binary image

Fig. 6 Using blob static analysis cropped image

107

108

M. Hanumanthu et al.

Fig.7 Bounding boxes on the image

Fig. 8 Final output of the image

5 Conclusion Extensive experimentation on a variety of images is shown, as is the performance evaluation of the proposed method. When we detect the face mask, the bounding box technique allows us to assist in identifying the object extremely precisely. The simplest technique presented in this study has been empirically confirmed for identifying the wearing of face masks and classifying the type of masks, which helps protect public health and plays a beneficial role in promoting the pandemic. Face restoration using PCA, SVM, and other machine learning algorithms that can recreate the face hidden behind the face mask could be the focus of future research in this field.

Facemask Detection Using Bounding Box Algortihm Under COVID-19 …

109

References 1. Maurya, S., Joseph, S., Asokan, A., Algethami, A. A., Hamdi, M., & Rauf, H. T. (2021). Federated transfer learning for authentication and privacy preservation using novel supportive twin delayed DDPG (S-TD3) algorithm for IIoT. Sensors, 21, 7793. 2. Putri, S. N. (2023). Application of fuzzy inference system mamdani at pelican crossing. In Proceedings of the International Conference on Rehabilitation and Maintenance in Civil Engineering, Surakarta, Indonesia (pp. 681–691). Springer: Singapore. Retrieved July 8–9, 2021 3. Ren, S., He, K., Girshick, R., & Sun, J. (2016). Faster r-cnn: Towards real-time object detection with region proposal networks. IEEE Transactions on Pattern Analysis and Machine Intelligence, 39(6), 1137–1149. 4. Karimullah, S., & Vishnu Vardhan, D. (2022). Pin density technique for congestion estimation and reduction of optimized design during placement and routing. Applied Nanoscience. 5. Balakrishna, S., Solanki, V. K., Gunjan, V. K., & Thirumaran, M. (2020). Performance analysis of linked stream big data processing mechanisms for unifying IoT smart data. In V. Gunjan, V. Garcia Diaz, M. Cardona, V. Solanki, & K. Sunitha (Eds.), ICICCT 2019—System Reliability, Quality Control, Safety, Maintenance and Management. Singapore: Springer. https://doi.org/ 10.1007/978-981-13-8461-5_78 6. Ruiz, V., Sánchez, Á., Vélez, J. F., & Raducanu, B. (2019). Automatic image-based waste classification. In Proceedings of the International Work-Conference on the Interplay between Natural and Artificial Computation, Almería, Spain (pp. 422–431). Retrieved June 3–7, 2019. 7. Jaya Krishna, N., Shaik, F., Harish Kumar, G. C. V., Naveen Kumar Reddy, D., & Obulesu, M. B. (2021). Retinal vessel tracking using gaussian and radon methods. In A. Kumar, & S., Mozar (Eds.), ICCCE 2020 (Vol. 698). Lecture notes in electrical engineering. Singapore: Springer. https://doi.org/10.1007/978-981-15-7961-5_37 8. Karimullah, S., Basha, S. J., Guruvyshnavi, P., Sathish Kumar Reddy, K., & Navyatha, B. (2020). A genetic algorithm with fixed open approach for placements and routings. In ICCCE (pp. 599–610). Publisher Springer. 9. Gunjan, V. K., Shaik, F., & Kashyap, A. (2021). Detection and analysis of pulmonary TB using bounding box and K-means algorithm. In A. Kumar, S. Mozar (Eds.), ICCCE 2020 (Vol. 698). Lecture notes in electrical engineering. Singapore: Springer. https://doi.org/10.1007/978-98115-7961-5_142 10. Saha, H. N., Auddy, S., Pal, S., Kumar, S., Pandey, S., Singh, R., Singh, S. K., Banerjee, S., Ghosh, D., & Saha, S. (2017). Waste management using the Internet of Things (IoT). In Proceedings of the 2017 8th Annual Industrial Automation and Electromechanical Engineering Conference (IEMECON), Bangkok, Thailand (pp. 359–363). Retrieved August 16–18, 2017. 11. Rashid, E., Ansari, M. D., Gunjan, V. K., & Ahmed, M. (2020). Improvement in extended object tracking with the vision-based algorithm. In V. Gunjan, J. Zurada, B. Raman, & G. Gangadharan (Eds.), Modern approaches in machine learning and cognitive science: A walkthrough. Studies in computational intelligence (Vol. 885). Springer, Cham. https://doi.org/10.1007/978-3-03038445-6_18 12. P. Sermanet, D. Eigen, X. Zhang, M. Mathieu, R. Fergus, & Lecun, Y. (2014). OverFeat: Integrated recognition, localization and detection using convolutional networks. 13. Karimullah, S., Vishnu Vardhan, D., & Basha, S. J. (2020). Floorplanning for placement of modulesin VLSI physical design using harmony search technique. In ICDSMLA 2019. Lecture notes in electrical engineering (Vol. 601). Springer Nature Singapore Pte Ltd.

110

M. Hanumanthu et al.

14. Nanni, L., Ghidoni, S., & Brahnam, S. (2017). Handcrafted versus non-handcrafted features for computer vision classification. Pattern Recognition, 71:158–172. https://doi.org/10.1016/j. patcog.2017.05.025 15. Nguyen, T. D., Rieger, P., Miettinen, M., & Sadeghi, A. -R. (2020). Poisoning attacks on federated learning-based IoT intrusion detection system. In Proceedings of the 2020 Workshop on Decentralized IoT Systems and Security, San Diego, CA, USA (pp. 1–7). Retrieved February 23–26, 2020. 16. Xiong, Z., Wang, Z., Du, C., Zhu, R., Xiao, J., & Lu, T. (2018). AnAasian face dataset and how race influences face recognition. In Pacific Rim Conference on Multimedia (pp. 372–383).

Accelerated Addition in Resistive Ram Array Using Parallel-Friendly Majority Gates J. Chinna Babu, Y. Suresh, R. Sudha Rani, S. Yasmeen, K. Siva Rama Krishna Reddy, and K. Harshavardhan

1 Introduction Hard lithography, Short-channel effects, and a considerable rise in power dissipation are all issues for CMOS-based systems [1]. QCA is among the most significant nanotechnologies proposed as a potential replacement for CMOS devices [2, 3]. Lenten et al. [4] proposed the concept of QCA in 1993. QCA has been physically implemented using four models: magnetic, Metal Island, molecular, as well as semiconductor [1], & numerous studies have lately described novel implementation breakthroughs [5, 6]. The fundamental building components of QCA are indeed an inverter as well as a majority voter. Many studies proposed this strategy for constructing a new majority gate configuration. The development of an appropriate majority gate structure enhance the effectiveness of the QCA circuit. Other strategies for optimising QCA circuits have been proposed, including such [7, 8]. Memory architecture piqued the attention of academics, particularly in QCA. QCA, like VLSI, includes a plethora of factors for evaluating circuit performance, such as latency, power dissipation, as well as area. The dependability of QCA circuits seems to be very crucial and must be carefully studied [9]. This study suggested a novel majority voter structure having three inputs and used it to build a new reduced power RAM cellular components. A power dissipation study is also given. The suggested architecture offers many advantages, including decreased power consumption, a small footprint, and single-layer implementations. Furthermore, the inputs are located outside the design, avoiding J. C. Babu (B) · Y. Suresh · R. S. Rani · S. Yasmeen · K. S. R. K. Reddy · K. Harshavardhan Department of Electronics and Communication Engineering, Annamacharya Institute of Technology and Sciences, Rajampet 516126, India e-mail: [email protected] Y. Suresh Department of Information Technology, PVP Siddhartha Institute of Technology, Vijayawada, India © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 V. K. Gunjan et al. (eds.), Modern Approaches in Machine Learning and Cognitive Science: A Walkthrough, Studies in Computational Intelligence 1117, https://doi.org/10.1007/978-3-031-43009-1_10

111

112

J. C. Babu et al.

crosstalk, making the gate more expandable and durable. In this study, the principles of QCA are reviewed in Sect. 2, the current SRAM is evaluated in Sect. 3, the suggested designs are examined in Sect. 4, and also the simulated results as well as conclusion are presented in Sects. 5 and 6, consecutively.

2 QCA Overview QCA circuits are quite often constructed utilizing QCA cells, that are similar elements. The QCA cell, as shown in Fig. 1, is formed like that of a square having four dots at every corner containing two charged electrons which may freely travel across them. The coulombic force causes electrons to move towards the dots on directly opposite corners. Each cell is rather different since it carries the right quantity of charge. The quantum dot appers to be a nm size semi—conducting nanostructure component with a 2–10 nm diameter. Valence electrons may move to other quantum dots via tunneling through QCA cell [10, 11]. As a result of columbic reactions, two different polarizations may develop. Every cell retains two extra free electrons within antipodal zones owing to Coulombic forces. As little more than a result, the suitable states ‘0’ as well as ‘1’ each have two alternative ground stages having linear polarisation coefficients of ‘1’ as well as ‘−1’, accordingly. The inverter & majority gate constitute two crucial parts of each and every QCA circuit. In actuality, the two most prominent logic gates within QCA are indeed an inverter as well as a three-majority gate. Figure 2 illustrates an inverter design, while Fig. 3 depicts a three-majority gate. Using QCA technology, all logical operations are performed using an inverter as well as a three-majority. If the inverter gate or even the three majority gates were previously used as an important gate throughout circuit design, study has shown that three majority gate generally underused in technology mapping. Numerous scholars are still researching novel logic gates in order to improve circuit efficiency [12, 13]. The functioning of this gate is shown in Eq. (1) Maj (A, B, C) = AB + AC + BC. The polarization of the central cell is determined by the 3- input cells of the Majority Voter, as well as the output received is the Majority Vote Polarity. Majority Fig. 1 Basic QCA cell

Accelerated Addition in Resistive Ram Array Using Parallel-Friendly …

113

Fig. 2 QCA Inverter

Fig. 3 Majority three input gate

gate is indeed the universal gate of a QCA Logic. It would be used as an Or even OR gate by reversing the polarity from one of the inputs to +1 or −1. In the provided function, the polarity of the input is set at +1 and −1. M (A, B, 0) = F = A B

(1)

M (A, B, 1) = F = A + B

(2)

Using the Three Input Majority Gate, several researchers have constructed both Combinational and Sequential Circuits. The QCA Designer programme was used to develop and evaluate the QCA structures using two distinct simulation engines [14–16]. Figures 4, 5, 6 show implementations of 3-input majority gates, or & and gates.

114 Fig. 4 Implementation of 3input majority gate

Fig. 5 Implementation of OR gate

Fig. 6 Implementation of AND gate

J. C. Babu et al.

Accelerated Addition in Resistive Ram Array Using Parallel-Friendly …

115

Fig. 7 Basic RAM cell design

3 Existing RAM Cell Developing the RAM cell would be a key issue in QCA Nano-technique. Generally practice, there have been two mechanisms for storing bits in QCA technology: loopbased as well as line-based. Moving the bit back across four time zones would retain it in a loop-based method. The prior value, but in the other side, was stored through a process known as line-based by employing the QCA line design. Figure 7 shows the fundamental RAM cell schematic diagram [15, 16].

4 Proposed RAM Cell A majority gate would be a customised gate that generates output that is comparable to the majority of inputs logic levels. Unless the majority of the gate’s sources are ‘1’, then the majority’s out is ‘1’; otherwise, the out is 0. The logic architecture for the majority gate-based memory module is depicted in Fig. 8. The developed memory cell consists primarily of a write/read, a select, an input, as well as an output signals. If utilised in a wide array, a select line would be used as a row /column selection. To pick a cell, connect the row/column decoder output to the selection line. If Select = ‘0’, the output is ‘0’ regardless of other input signals. This indicates that perhaps the cell could not be used for memory operations and also that the circuitry would be in the hold mode. If Select = ‘1’, the write and read operations are carried out depending on the Write/Read’ inputs. If Write/ Read’ equals to ‘1’, then Write operation is carried out. Unless the input is ‘1’, the Q becomes ‘1’ since the majority of the input to the majority gate represent ‘1’, and vice versa for writing ‘0’. Throughout a read operation, the Write/Read’ equals ‘0’ as well as the Q would be in hold mode, reading the previously stored data. Figure 9

116

J. C. Babu et al.

Fig. 8 Proposed RAM cell using 3-input majority gate cell

depicts the configuration of the Majority circuit SRAM built using QCA technology. To get the entire Memory cell architecture, many fundamental gates such as AND, OR, NOT, and Majority gates are created and combined. All of the cells’ clocking has indeed been designed in a manner that we obtain undistorted output. Fig. 9 QCA based proposed SRAM

Accelerated Addition in Resistive Ram Array Using Parallel-Friendly …

117

5 Simulation Results To construct the SRAM cell, a Majority gate was integrated into the design. A QCA Designer-E tool was used to create the Majority circuit SRAM cell, as well as its output Read/Write functionalities were confirmed. Write operation: The selection line & Write/Read inputs are maintained high during Writing operation. The output is given or low depending on the input value. Unless the input would be a ‘1’, the majority of an input to a Majority gate are also ‘1’, and the output would be high. Likewise, if indeed the input equals ‘0’, the Majority gate’s majority inputs are low, and so the out is low. Clock 1 is acquiring the output. Read operation: The Write/Read inputs is maintained low throughout the Read operation, while the Selection line is driven up. The output would follows the memory loop Q bit regardless of the value of the parameter. Figure 10 depicts the output operation. The comparison of Basic SRAM and Majority-Gate based SRAM is shown in Table 1. Performance/energy measurements have been used to assess the performance of SRAMs based on Basic & Majority Gates. As compared to regular SRAM, MajorityGate SRAM consumes less power as well as energy.

Fig. 10 Simulation result of proposed RAM cell

Table 1 Shows a comparison between Fundamental SRAM versus Majority-Gate driven SRAM

BASIC RAM

Proposed

Power

26 pW

20 pW

Energy

2.734 e–2 eV

2.715 e–2 eV

118

J. C. Babu et al.

6 Conclusion We present a new majority-3 gate configuration. The majority gate would be used to create a one-of-a-kind QCA-RAM cell with power to set or reset its output. The QCADesigner tool would be used for circuit design and implementation in this work, whereas the QCAPro programme is used for power analysis. The proposed structure have always had the following benefits: compact size, low consumed power, low cost, plus simple implementation. The recommended circuits are much more realistic for physical implementation since they are constructed on a single layer.

References 1. Khosroshahy, M. B., Moaiyeri, M. H., Navi, K., & Bagherzadeh, N. (2017). An energy and cost efficient majority-based RAM cell in quantum-dot cellular automata. Results in Physics, 7, 3543–3551. 2. Gunjan, V. K., Prasad, P. S., Fahimuddin, S., & Bigul, S. D. (2019). Experimental investigation to analyze cognitive impairment in diabetes mellitus. In A. Kumar, & S. Mozar (Eds.), ICCCE 2018 (Vol. 500). Lecture notes in electrical engineering. Singapore: Springer. https://doi.org/ 10.1007/978-981-13-0212-1_79 3. Pan, J., Chang, C. C., Xie, Z., Li, A., Tang, M., Zhang, T., Hu, J., Chen, Y. (2022). Towards collaborative intelligence: Routability estimation based on decentralized private data. In Proceedings of the DAC ’22, 59th ACM/IEEE Design Automation Conference, San Francisco, CA, USA (pp. 961–966). Retrieved 10–14 July, 2022. 4. Sheikhfaal, S., Angizi, S., Sarmadi, S., Moaiyeri, M. H., & Sayedsalehi, S. (2015). Designing efficient QCA logical circuits with power dissipation analysis. Microelectronics Journal, 46(6):462–471. 5. Karimullah, S., Vishnuvardhan, D., & Bhaskar, V. (2022). An improved harmony search approach for block placement for VLSI design automation. Wireless Personnel Communications, 127, 3041–3059. https://doi.org/10.1007/s11277-022-09909-2 6. Jaya Krishna, N., Shaik, F., Harish Kumar, G.C.V., Naveen Kumar Reddy, D., Obulesu, M.B. (2021). Retinal Vessel Tracking Using Gaussian and Radon Methods. In: Kumar, A., Mozar, S. (eds) ICCCE 2020. Lecture Notes in Electrical Engineering, vol 698. Springer, Singapore. https://doi.org/10.1007/978-981-15-7961-5_37. 7. Nguyen, D. C., Ding, M., Pathirana, P. N., Seneviratne, A., Li, J., & Poor, H. V. (2021). Federated learning for internet of things: A comprehensive survey. IEEE Communications Surveys and Tutorials, 23, 1622–1658. 8. Chakraborty, S., Onuchowska, A., Samtani, S., Jank, W., & Wolfram, B. (2021). Machine learning for automated industrial IoT attack detection: An efficiency-complexity trade-off. ACM Transactions on Management Information Systems, 12, 1–28. 9. Karimullah, S., & Vishnuvardhan, D. (2020). Experimental analysis of optimization techniques for placement and routing in Asic design. In ICDSMLA 2019 (Vol. 601). Lecture notes in electrical engineering. Springer Nature Singapore Pte Ltd. 10. Goswami, M., Sen, B., Mukherjee, R., & Sikdar, B. K. (2017). Design of testable adder in quantum-dot cellular automata with fault secure logic. 11. Agiollo, A., Conti, M., Kaliyar, P., Lin, T.-N., & Pajola, L. (2021). DETONAR: Detection of routing attacks in RPL-based IoT. IEEE Transactions on Network and Service Management, 18, 1178–1190.

Accelerated Addition in Resistive Ram Array Using Parallel-Friendly …

119

12. Microelectronics Journal 60:1–123. Bruschi, F., Perini, F., Rana, V., & Sciuto, D. (2011). An efficient quantum-dot cellular automata adder. In 2011 Design, Automation & Test in Europe (pp. 1–44). 13. Karimullah, S., & Vishnuvardhan, D. (2020). Iterative analysis of optimization algorithms for placement and routing in Asic design. In ICDSMLA 2019 (Vol. 601). Lecture notes in electrical engineering. Springer Nature Singapore Pte Ltd. 14. Gonelli, M., Fin, S., Carlotti, G., Dey, H., Csaba, G., Porod, W., et al. (2018). Robustness of majority gates based on nanomagnet logic. Journal of Magnetism and Magnetic Materials, 460, 432–437. 15. Das, J. C., De, D., Mondal, S. P., Ahmadian, A., Ghaemi, F., & Senu, N. (2019). QCA based error detection circuit for nano communication network. IEEE Access, 7, 67355–67366. 16. Saha, H. N., Auddy, S., Pal, S., Kumar, S., Pandey, S., Singh, R., Singh, S. K., Banerjee, S., Ghosh, D., & Saha, S. (2017). Waste management using the Internet of Things (IoT). In Proceedings of the 2017 8th Annual Industrial Automation and Electromechanical Engineering Conference (IEMECON), Bangkok, Thailand, (pp. 359–363). Retrieved August 16–18, 2017.

Optimization of Area and Wirelength Using Hybrid BPSO Algorithm in VLSI Floorplan and Placement for IC Design Shaik Karimullah, D. Vishnuvardhan, Vinit Kumar Gunjan, and Fahimuddin Shaik

1 Introduction As the market for cutting-edge technology grows, more research is being done in the field of IC design, resulting in VLSI designs that are more composite, reliable, compact, and perform better [1, 2]. The amount of transistors in a single chip has increased significantly, which has greatly complicated IC design. Floorplanning, a fundamental physical design phase for hierarchical, building-block design techniques [3, 4], is a more sophisticated solution to the aforementioned challenge. It is challenging to determine the global solution space since the solution space expands together with the size of the circuit. The Hybrid BPSO technique can be used to handle the problem of floorplanning [5].

1.1 Background It is difficult to notice early feedback in the context of IC physical design in order to predict the area, latency, wire length, and so on prior to IC manufacture [6].

S. Karimullah (B) · F. Shaik Department of ECE, AITS, Rajampet, India e-mail: [email protected] D. Vishnuvardhan Department of ECE, JNTUACE, Ananthapuramu, India V. K. Gunjan Department of Computer Science and Engineering, CMR Institute of Technology Hyderabad, Hyderabad, India © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 V. K. Gunjan et al. (eds.), Modern Approaches in Machine Learning and Cognitive Science: A Walkthrough, Studies in Computational Intelligence 1117, https://doi.org/10.1007/978-3-031-43009-1_11

121

122

S. Karimullah et al.

1.2 Objectives There is a significant difficulty in the arena of IC physical design to discover early feedback to predict the area, wire length, latency, and so on before IC manufacture [7].

1.3 Need and Importance • According to Moore’s law, the number of components on modern ICs has increased around two times every two years. • As a result, producing an optimal floorplan becomes impossible and must rely on certain valuable algorithms.

2 Literature Review The authors of [8] discuss PSO using a polish expressing themselves to fix modules alongside a fixed outline constraint and a hybrid ant colony optimisation approach which may reduce computation time as well as provide better floor-plan arrangements. The latter approach focuses on creating a more suitable floor-plan in less time. To reach both a global and local optimum, [9] offer a hybrid strategy combining PSO with Harmony Search (HS). The findings indicate that there is an ideal solution for a larger number of modules. Reference [10] To explore the search space and find a workable solution, the researcher offers a discrete PSO (DPSO) technique. For non-slicing floor layouts, [11] propose a hybrid particle swarm and ant colony optimisation (PSO/ACO) strategy to obtain the optimal result.

2.1 Drawbacks • The PSO method has the drawback of having a sluggish rate of convergence during repeated processes and being easy to slip towards a local optimum within a high-dimensional environment [12]. • It generates low-quality solutions and necessitates the use of memory to update velocity [13]. • If the problem is discrete, PSO isn’t as effective [14].

Optimization of Area and Wirelength Using Hybrid BPSO Algorithm …

123

3 Methodology 3.1 Existing Method The fundamental PSO is a search method based on a stochastic population. It is considered to be a workable option to the majority of challenging non-linear optimisation problems. It was initially introduced by Drs. Kennedy and Eberhart in 1995. The comparison is based on a flock of birds and insects moving about in a search space for search of food. These birds and insects do not know where the best location is inside the search space [15–18]. Through their social conduct, one individual can show the other group members a desired way, and they will quickly copy them. To address optimization problems, this approach is derived from animal behaviour or activity [19–21]. Particles in the PSO swarm communicate with each other to find suitable places. Based on the optimal position of every particle in the swarm, they adjust their positions and speeds in real time. All particles throughout the search space look for better locations up until the swarm gets close to the optimal fitness function. The PSO algorithm’s success can be ascribed to its simplicity in use and capacity to converge on a better solution. It doesn’t consider the gradient information of the function; it only uses mathematical operators. The basic PSO is not much effective when it is discrete in nature [22–24].

3.2 Proposed Method In the proposed methodology, combing of both PSO and BAT algorithms properties will be done to provide better optimal solutions which can be termed as Hybrid Bat & Particle Swarm Optimization algorithm. The underlying PSO algorithm motivates the proposed optimisation approach. In this example, the population is represented by the number of pieces to be put in a floor plan space. All parameters, including optimisation iterations, population size, learning coefficients and inertia weights must be loaded at the start. The modules were rectangular in shape and vary in size; both the height and width of each module must be determined prior to the optimisation process. Because each particle travels at a different speed, a velocity matrix has exactly the same amount of rows and columns that the number of modules. Swap, rotate, and move are the three major operations performed by the PSO on the particles/modules [25, 26]. When two particles exchange locations, they do so with the idea that the new position is the best fit for them. If the optimum position may be occupied, the individual particle must alter its position or rotate at right angles in order to move and rotate. Until all of the particles are in the ideal location or the termination condition is reached, all of the particles conduct these three processes as shown in Fig. 1. The number of iterations is the termination condition.

124

S. Karimullah et al.

Fig. 1 Flowchart of BPSO

The typical BA involves four basic parameters: pulse frequency, pulse rate, velocity, and a constant. There is no limit to variegate the BA over mutation acceptance, parameter assortment, or hybridizing the bat algorithm with other algorithms to improve the local search capabilities and maintain population divergence so that the algorithm avoids trapping in local optima. Although, the problem with the BA modification method is that it never ensures that a modified BA finds a worldwide optimum solution. Hence by combining the properties like pulse frequency, pulse rate, velocity, swapping, rotation resulting in the Hybrid BPSO Algorithm which produces very optimal solutions compared to other algorithms.

Optimization of Area and Wirelength Using Hybrid BPSO Algorithm …

3.2.1

125

Flow Chart

4 Implementation Tool For the purposes of this study, we will do image processing calculations using the MATLAB application. A sophisticated language for programming used for specialist registers is called MATLAB. It combines computation, representation, and coding in an approachable way, expressing disagreements and agreements through organic numerical documentation. The following activities are also carried out using MATLAB: Algorithm progress. • • • •

Math and calculation Data collection Modeling, reenactment, and prototyping Data analysis, inquiry, and representation.

5 Results and Discussion MATLAB R2017a is used to implement the HPSO algorithm for VLSI floorplan optimization. Set the number of blocks and the maximum number of iterations required for the optimization. For each new optimization problem, certain parameters must be defined, such as the population or swarm size, the width and height of the parent module (W and H), the width and height of modules (w and h), random points on modules (rin and rout), spacing between modules (d), and the velocity matrix, which is a n*n matrix where n corresponds to the number of particles.

5.1 Output 1 Here initially we provided with 50 number of blocks where these undergoes 150 number of iterations. At the initial stage these blocks were placed randomly. But after implementation of Hybrid BPSO these are getting modified by each iteration as it undergoes. By the completion of all the provided iterations resulting in the minimization of area and wirelength which is as shown in the above Fig. 2.

5.2 Output 2 Provided with 70 number of blocks and undergoes 500 number of iterations. As these no of blocks are situated at different positions before implementation of Hybrid BPSO algorithm, modified continuously until these situated best position among

126

S. Karimullah et al.

Fig. 2 Simulation results for 50 blocks with 150 iterations

Fig. 3 Simulation results for 70 blocks with 500 iterations

all the locations resulting in optimal minimization of area, wirelength and delay as shown in Fig. 3.

5.3 Output 3 In this case we are provided with 100 number of blocks along with 150 number of iterations. Beforehand these modules were randomly positioned and after implementation of Hybrid BPSO algorithm which alternates their positions finally resulting in the overall reduction of dead space, area, wirelength, delay as shown in Fig. 4.

Optimization of Area and Wirelength Using Hybrid BPSO Algorithm …

127

Fig. 4 Simulation results for 100 blocks with 150 iterations

As compared to other algorithms, the proposed Hybrid BPSO algorithm produces better optimal results and comparative results are as shown in Tables 1, 2 and the graphical relation is shown in Fig. 5. Table 1 The details of MCNC benchmark circuits Problems

Modules

Nets

Pins

Terminals

Total area of modules

Apte

9

97

287

73

46.56

Xerox

10

203

698

2

19.35

Hp

11

83

309

45

8.83

Ami33

33

123

522

42

1.15

Ami49

49

408

953

22

35.44

Table 2 Comparison of results Apte Area

Xerox Wire

Area

Hp Wire

Area

Ami33 Wire

Area

Ami49 Wire

Area

Wire

GA-Otree

47.4

263

20.1

462

9.4

152

1.23

49.2

37.8

725

HPSOHS

46.9

191

20.2

500

9.85

68.3

1.29

46.2

39.5

912

HPSOFF

47.4

263

20.1

481

9.46

153

1.23

57.2

38.5

689

PSO

47.3

263

20.2

477

9.5

136

1.28

69

38.8

880

HBPSO algorithm

45.6

260

20.0

456

9.2

122

1.21

45.3

36.5

732

128

S. Karimullah et al.

1000 900 800 700 600 500 400 300

200 100 0 Area

Wire Apte

Area

Wire Xerox

GA-Otree

HPSOHS

Area

Wire Hp

HPSOFF

Area

Wire Ami33

PSO

Area

Wire Ami49

HBPSO Algorithm

Fig. 5 Graphical relationship among various optimization algorithms and BPSO

5.4 Comparative Results 6 Conclusion A Hybrid BPSO technique for VLSI floor-plan optimization is implemented in this study. The results of the experiments suggest that Hybrid BPSO is the optimum floorplan optimization approach. When compared to other existing approaches, it gives better outcomes. To test the viability, the implementation was first carried out on a few modules with a smaller floor-plan space. In comparison to the initial floorplan area, an improvement in area was gained for floor-plans with a certain number of modules. The elapsed time is long because the number of iterations was set to 1000.

References 1. Xie, Z., Pan J., Chang, C. C., Liang, R., Barboza, E. C., & Chen, Y. (2022). Deep learning for routability. In H. Ren, & J. Hu (Eds.), Machine learning applications in electronic design automation (pp. 35–61). Cham, Switzerland: Springer International Publishing. ISBN 978-3031-13074-8 2. Yang, X. S. (2010). A new metaheuristic bat-inspired algorithm. In Nature inspired cooperative strategies for optimization (NICSO 2010) (pp. 65–74). Berlin, Heidelberg: Springer. 3. Lakshmanna, K., Shaik, F., Gunjan, V. K., Singh, N., Kumar, G., & Shafi, R. M. (2022) Perimeter degree technique for the reduction of routing congestion during placement in physical design of VLSI circuits. Complexity, 2022:11. https://doi.org/10.1155/2022/8658770 4. Weng, Y., Chen, Z., Chen, J., & Zhu, W. (2019). A modified multi-objective simulated annealing algorithm for fixed-outline floor-planning. In IEEE International Conference on Automation Electronics and Electrical Engineering Electronic. ISBN: 978-1- 5386-7861-9.

Optimization of Area and Wirelength Using Hybrid BPSO Algorithm …

129

5. Shaik Karimullah, Vishnuvardhan, D., Muhammad, A., Gunjan, V. K., & Fahimuddin, S. (2022). Kazy Noor-e-alam Siddiquee, An improved harmony search approach for block placement for VLSI design automation. Wireless Communications and Mobile Computing, 2022, Article ID 3016709, 10. https://doi.org/10.1155/2022/3016709 6. Subbaraj, P., Siva Kumar, P., & Pandiraj (2008). Memetic algorithm for solving channel routing problems. In International Conference on VLSI and Signal Processing (ICVSP ‘08). Chennai, India. 7. Karimullah, S., Vishnu Vardhan, D. (2022). Pin density technique for congestion estimation and reduction of optimized design during placement and routing. Applied Nanoscience. 8. Lee, K. Y., & Al-Sharkawi, M. (2008). Modern heuristic optimization techniques: Theory and application to power systems. In Fundamentals of Particle Swarm Optimization Techniques Willey-Interscience Hoboken (pp. 72–79). 9. Ingber, L. (1989). Very fast simulated re-annealing. Mathematical and Computer Modelling, 12(8), 967–973. 10. Rashid, E., Ansari, M. D., Gunjan, V. K., & Ahmed, M. (2020). Improvement in extended object tracking with the vision-based algorithm. In: Modern approaches in machine learning and cognitive science: A walkthrough: Latest trends in AI (pp. 237–245). 11. Kim, D., Kwon, H., Lee, S. Y., Kim, S., Woo, M., & Kang, S. (2021). Machine learning framework for early routability prediction with artificial netlist generator. In Proceedings of the 2021 Design, Automation & Test in Europe Conference & Exhibition (DATE), Grenoble, France (pp. 1809–1814). Retrieved February 1–5, 2021. 12. Shanavas, H., & Gnanamurthy, R. K. Evolutionary algorithmical approach on VLSI floorplanning problem. International Journal of Computer Theory and Engineering, 1(4), 461–464. 13. Cincotti, A., Cuttelo, V., & Pavone, M. (2002). Graph partitioning using genetic algorithms with OPDX. In Proceedings of IEEE on World Congress on Computational Intelligence (pp. 402– 406). 14. Nath, R. (2022). Role of the fourth industrial revolution towards sustainable development. In Evolution of digitized societies through advanced technologies (pp. 131–137). Singapore: Springer Nature Singapore. 15. Xu, N., Huang, F., Jiang, Z. (2006). A fast algorithm for VLSI building block placement. In Proceedings of the IMACS Multiconference on “Computational Engineering in Systems Applications” (CESA ‘06) (pp. 2157–2160), Beijing, China. 16. Gunjan, V. K., Shaik, F., Kashyap, A. (2021). Detection and analysis of pulmonary tb using bounding box and K-means algorithm. In A. Kumar, & S. Mozar (Eds.), ICCCE 2020 (Vol. 698). Lecture notes in electrical engineering. Singapore: Springer. https://doi.org/10.1007/978981-15-7961-5_142 17. Pan, J., Chang, C. C., Xie, Z., Li, A., Tang, M., Zhang, T., Hu, J., & Chen, Y. (2022). Towards collaborative intelligence: Routability estimation based on decentralized private data. In Proceedings of the DAC ’22, 59th ACM/IEEE Design Automation Conference, San Francisco, CA, USA (pp. 961–966). Retrieved July 10–14, 2022. 18. Thakral, M., Singh, R. R., & Singh, S. P. (2022). An extensive framework focused on smart agriculture based out of IoT. In Evolution of digitized societies through advanced technologies (pp. 139–152). Singapore: Springer Nature Singapore. 19. Yılmaz, S., UgurKucuksille, E., & Cengiz, Y. (2014). Modified bat algorithm (Vol. 20, pp. 71– 78). 20. Karimullah, S., Vishnu Vardhan, D., & Basha, S. J. (2020). Floorplanning for placement of modules in VLSI physical design using harmony search technique. In ICDSMLA 2019 (Vol. 601). Lecture notes in electrical engineering. Springer Nature Singapore Pte Ltd. 21. Kahng, A. B. (1995). INVITED: Reducing time and effort in IC implementation: A roadmap of challenges and solutions. In Proceedings of the Proceedings of the 55th Annual Design Automation Conference, San Francisco, CA, USA (pp. 1–6). Retrieved June 24–29, 2018; Kennedy, J., & Eberhart, R. C. (1995). Particle swarm optimization. In Proceedings of the IEEE Int’l Conference on Neural Networks (pp. 1942–1948).

130

S. Karimullah et al.

22. Srinivas, B., Satheesh, P., & Rama Santosh Naidu, P. (2022). Surface damage indicators identification through simulation studies in computer vision. In Proceedings of the International Conference on Computer Vision, High Performance Computing, Smart Devices and Networks: CHSN-2020 (pp. 131–137). Singapore: Springer Nature Singapore. 23. Chang, Y. C., Chang, Y. W., Wu, G. M., & Wu, S. W. (2000). B*-trees: A new representation for non-slicing floorplans. In Proceedings of the DAC (pp. 458–463). 24. Dorigo, M., Maniezzo, V., & Colorni, A. (1996). Ant system: Optimization by a colony of cooperating agents. IEEE Transactions on Systems Man and Cybernetics -Part B, 26(1), 29–41. 25. Agiollo, A., Conti, M., Kaliyar, P., Lin, T.-N., & Pajola, L. (2021). DETONAR: Detection of routing attacks in RPL-based IoT. IEEE Transactions on Network and Service Management, 18, 1178–1190. 26. Venkatesan, M., Kulkarni, A., & Menon, R. (2020). Marginal deep architectures based massive MIMO beam forming and predictions. Helix—The Scientific Explorer | Peer Reviewed Bimonthly International Journal, 10(04), 26–31. Retrieved from https://helixscientific.pub/ index.php/home/article/view/147

PAPR and SER Performance Analysis of OFDMA and SCFDMA G. Obulesu, Shaik Karimullah, Fahimuddin Shaik, M. Nanda Krishna, C. Pavan Kumar, G. Divyanjali, and S. Mohammad Anas

1 Introduction Mobile communication has rapidly increased because of the demand of high data rates and throughputs. In order to achieve error-free high bit rates, multi-carrier transmission methods are utilised in current wireless mobile systems [1]. The application of multi-carrier modulation (MCM) has developed into a common method for accomplishing this goal. OFDM, which stands for orthogonal frequency division multiplexing, is the most well-known example of MCM transmission [2]. The common orthogonal frequencies-division multiplexing (OFDM) electronic modulation technique is a multi-user orthogonal frequency-division multiple access (OFDMA). OFDMA provides multiple access by allocating individual users portions of subcarriers. This enables multiple users to send low-data-rate messages at the same time [3]. However, both OFDM and OFDMA suffer from power distortion, which can be especially problematic in uplink transmissions. Because of its lower power distortion and lower BER and SER, single-carrier frequency division multiple access (SC-FDMA) was chosen as the uplink multiple access scheme to overcome this limitation [4]. SC-FDMA is used because its peak-to-average-power-ratio (PAPR) is smaller and more constant. Furthermore, it has a comparable throughput and, for the most part, the same overall complexity as an orthogonal frequency division multiple access (OFDMA) system [5].

G. Obulesu (B) · S. Karimullah · F. Shaik · M. N. Krishna · C. P. Kumar · G. Divyanjali · S. M. Anas Electronics and Communication Engineering, Annamacharya Institute of Technology and Sciences, Rajampet, India e-mail: [email protected] F. Shaik e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 V. K. Gunjan et al. (eds.), Modern Approaches in Machine Learning and Cognitive Science: A Walkthrough, Studies in Computational Intelligence 1117, https://doi.org/10.1007/978-3-031-43009-1_12

131

132

G. Obulesu et al.

Single-carrier FDMA (SC-FDMA) is a frequency-division multi-access system. It’s sometimes referred to as linear precoded OFDMA (LP-OFDMA). It, like other multiple access methods (TDMA, CDMA, FDMA, and OFDMA), deals with the assignment of numerous users to a common communication resource. Because it adds an extra DFT processing step preceding the standard OFDMA processing, SC-FDMA may be regarded of as a linear precoded OFDMA scheme [6]. SC-FDMA has gained traction as a viable alternative to OFDMA, especially in upstream communications where a reduced peak-to-average power ratio (PAPR) helps the mobile device in terms of transmission power consumption and power amplifier cost [7]. It is utilised in 3GPP Long Term Evolution (LTE), additionally referred to as Evolved UTRA (E-UTRA) [8], as the uplink multiple access technique. Several studies have been conducted on the performance of SC-FDMA in comparison to OFDMA. Despite the small performance difference, Because of its low PAPR, SC-FDMA is ideal towards uplink wireless data transfer in mobile telephone networks wherein transmitter power consumption is crucial.

2 Literature Review This study compares two single-carrier frequency-division multiple access (SCFDMA) methods for a satellite uplink, namely localised FDMA (LFDMA) and interleaved FDMA (IFDMA) to orthogonal FDMA (OFDMA). The latter uses the digital television broadcasting (DVB) standards family for its air interface [9]. Using two cutting-edge high power amplifiers (HPAs) working in the K- and S-bands, the effectiveness of both asynchronous as well as synchronous LFDMA, IFDMA, and OFDMA within a multi-user scenario is examined. While IFDMA beats the other two systems for synchronous receiving, it is the most vulnerable to deterioration resulting from inter-block interference (IBI) for asynchronously reception (IBI) [10]. Furthermore, because to its relatively significant envelope variations, OFDMA is the most prone to non-linear distortion [11]. Although LFDMA beats IFDMA for synchronous receiving, it outperforms both systems for asynchronous reception, particularly with higher IBI distortion [12].

3 Methodology and Process Flow The transmission processing of SC-FDMA is quite similar to that of OFDMA. Each user’s bit sequence is mapped to a complicated constellation of symbols (BPSK, QPSK, or MQuadrature amplitude modulation) [13]. Then, various Fourier coefficients are allocated to various transmitters (users) [14]. The act of mapping then demapping blocks accomplish this process. The receiver side comprises one demapping block and one IDFT block, and a single detecting block for each user signals to be received. Guard intervals with cyclic repetition are provided among blocks

PAPR and SER Performance Analysis of OFDMA and SCFDMA

133

of symbols, exactly like in OFDM, to effectively minimise interference between symbols caused by temporal spreading (produced by multi-path propagation) within the blocks [15]. SC-FDMA allows for multiple user access by allocating multiple sets and non overlapping Fourier components to various users (sub-carriers) [16]. This is performed at the transmitter by injecting quiet Fourier coefficients (at places allotted to other users) before doing the IDFT and deleting them on the receiver side after the DFT. SC-FDMA is distinct from OFDMA in that it generates a single-carrier broadcast signal, whereas OFDMA uses multiple carriers [17]. Subcarrier mapping is classified into two types: localised mapping and spread mapping. The DFT outputs are assigned to a selection of successive subcarriers via localised mapping, restricting the bandwidth they use to a portion of the entire system’s bandwidth. The DFT output of the input information are non-continuously allocated to subcarriers throughout the whole bandwidth in dispersed mapping, leading to zero amplitude for each of the leftover subcarriers [18]. Interleaved SC-FDMA (IFDMA) is a distributed SC-FDMA variation whereby the occupied subcarriers are separated evenly throughout the whole bandwidth [19]. SC-FDMA offers a smaller peak-to-average power ratio (PAPR) over OFDM and OFDMA due to its intrinsic single carrier structure, allowing for more relaxed design requirements in a subscriber unit’s transmit channel. Unlike OFDM, which immediately modulates multiple subcarriers with transmit symbols, SC-FDMA broadcast symbols undergo initial processing through an N-point DFT blocks [20]. Following the DFT computation, equalisation is performed on the receiving side in OFDM and SC-FDMA through doubling each Fourier component by a complex integer. As a consequence, frequency-selective fading with phase distortion may be avoided with relative ease. The use of FFTs for bandwidth-domain equalisation has the benefit of requiring less computation than conventional time-domain equalisation, which necessitates multi-tap FIR or IIR-filters. With fewer calculations, there is less compounding round-off error, often known as numeric noise [21]. Combining a single carrier’s transmission with a single-carrier frequencydomainequalization (SC-FDE) technique is a similar concept. Single carrier transmission, unlike SC-FDMA and OFDM, does not employ IDFT or DFT at the transmitting device, instead applying the cyclic prefix to change the linear channel combination into a circular one. After removing the cyclic prefix from the receiver, a DFT is utilised to enter the domain of frequency, where a basic single-carrier frequencydomain-equalization (SC-FDE) method, following by the IDFT operation, can be applied. Figure 1 depicts an OFDMA and SC-FDMA block diagram. SC-FDMA adds one more module, DFT, before the IFFT component in the transmitter’s chain, and IDFT in the reception chain. This converts the OFDMA chain to a SC-FDMA chain. In the event of an absence of both of these modules, the chain is referred to as the OFDMA communication and reception chain. DFT outputs are allotted to M successive subcarriers from an aggregate of N subcarriers in LFDMA (where N>M). In contrast, DFDMA spreads the M amount

134

G. Obulesu et al.

Fig. 1 Model of OFDMA and SC-FDMA

of DFT outputs throughout the whole band. As illustrated in Figure 2, each DFDMA and LFDMA allocate null amplitude to (N-M) vacant subcarriers. The mapping of the mode is referred to as interleave FDMA (IFDMA) if all of the DFT output are dispersed with an equivalent distance N/M=Q among the occupied subcarriers. Q has been identified as the bandwidth spread factor.

Fig. 2 Example of different subcarrier mapping modes for M = 4, Q = 3 and N = 12

PAPR and SER Performance Analysis of OFDMA and SCFDMA

135

4 Results and Analysis The recommended techniques are put into action using the MATLAB computer-aided design tool. The tool version utilised in this study effort is r2017b, which includes the necessary communication along with signal processing toolboxes. SER versus SNR (in dB) for SC-FDMA System We have used the following data for our analysis: • • • • •

64-QAM modulated signals FFT/IFFT length = 512 Input-data block size = 32 Cyclic Prefix (CP) Length = 20 SNR ψ [0, 30 dB]

To evaluate the SER versus SNR (in dB) performance of SCFDMA System three different channels were used, which are pedAchannel, vehAchannel and AWGN channel. Furthermore, the plots were plotted for 3 different subcarrier mapping: localised FDMA (LFDMA), interleaved FDMA (IFDMA) and distributed FDMA (DFDMA). A. pedAchannel While using pedAchannel for SCFDMA, from Fig. 3 and Table 1, it is observed that the SER performance of LFDMA is better than IFDMA and DFDMA technique because of its robustness against multiple carrier interference. Also DFDMA and IFDMA are almost the same in their SER performance (and hence are overlapped when plotted in single figure).It is observed that the performance is greatly improved with SCFDMA as compared to OFDMA for all three subcarrier techniques in pedAchannel.

Fig. 3 SER versus SNR [dB] for SCFDMA in pedAchannel using LFDMA, IFDMA and DFDMA subcarrier mapping. (right side figure is added as a subplot to show the values as they are overlapped in left one)

136 Table 1 SER versus SNR [dB] for SCFDMA in pedAchannel

G. Obulesu et al.

SNR(db)

SER LFDMA

IFDMA

DFDMA

0

0.01

0.021

0.02

2

0.004

0.016

0.015

4

0.0001

0.0081

0.008

6

0.00002

0.0031

0.003

8

0.000003

0.00052

0.0005

10

0.000001

0.000061

0.00006

B. vehAchannel While using vehAchannel for OFDMA, from Fig. 4 and Table 2, it is observed that the SER performance of LFDMA is a lot better than IFDMA and DFDMA technique because of its robustness against multiple carrier interference. Also DFDMA and IFDMA are almost the same in their SER performance (and hence are overlapped when plotted in single figure). When compared to pedAchannel, the LFDMA performance is slightly better in vehAchannel but IFDMA and DFDMA performance is better for pedAchannel. When compared to vehChannel performance of OFDMA, a great improvement in SER performance is observed. Also, it must be noted that vehChannel performance of SCFDMA is better than the performance of OFDMA in all three channels. C. AWGN Channel While using the AWGN channel for OFDMA, from Fig. 5 and Table 3, it is observed that the SER performance of LFDMA, IFDMA and DFDMA is almost similar. When compared to pedAchannel, there is a reduction in SER performance of LFDMA in

Fig. 4 SER versus SNR [dB] for SCFDMA in vehAchannel using LFDMA, IFDMA andDFDMA subcarrier mapping. (right side figure is added as a subplot to show the values that are overlapped in left one)

PAPR and SER Performance Analysis of OFDMA and SCFDMA Table 2 SER versus SNR [dB] for SCFDMA in vehAchannel

SNR (db)

137

SER LFDMA

IFDMA

DFDMA

0

0.005

0.061

0.06

2

0.002

0.059

0.058

4

0.0003

0.055

0.054

6

0.000007

0.053

0.052

8

0.0000005

0.051

0.05

10

0.0000001

0.041

0.04

AWGN channel but SER performance IFDMA and DFDMA is better for AWGN channel. When compared to vehAchannel, there is a reduction in SER performance of LFDMA in AWGN channel but SER performance IFDMA and DFDMA is considerably better for AWGN channel. Also, it must be noted that the AWGN performance of SCFDMA is better than the performance of OFDMA in all three channels. PAPR Performance for SCFDMA System We have used the following data for our analysis: • 16-QAM modulated signals • FFT/IFFT length = 512 • Input-data block size = 32 To evaluate the PAPR performance of SCFDMA System three different subcarrier mapping were used, which are localised FDMA (LFDMA), interleaved FDMA (IFDMA) and distributed FDMA (DFDMA).

Fig. 5 SER versus SNR [dB] for SCFDMA in AWGN Channel using LFDMA, IFDMA and DFDMA subcarrier mapping. (right side figure is added as a subplot to show the values as they are overlapped in left one)

138 Table 3 SER versus SNR [dB] for SCFDMA in AWGN channel

G. Obulesu et al.

SNR (db)

SER LFDMA

IFDMA

DFDMA

0

0.02

0.025

0.026

2

0.015

0.0156

0.0157

4

0.005

0.0055

0.0056

6

0.0015

0.0016

0.0017

8

0.0002

0.00021

0.00022

10

0.00001

0.000011

0.000012

From Fig. 6 and Table 4, it is observed that the PAPR performance of LFDMA and DFDMA is almost similar in case of SCFDMA and IFDMA gives a better performance as compared to LFDMA and DFDMA in OFDMA. It was observed that the PAPR performance of SCFDMA significantly improved in all three subcarrier mapping techniques i.e LFDMA, IFDMA and DFDMA, when compared to their counterparts in OFDMA.

Fig. 6 CCDF versus PAPR for SCFDMA using LFDMA, IFDMA and DFDMA subcarrier

Table 4 CCDF versus PAPR for SCFDMA

PAPR (db)

CCDF LFDMA

IFDMA

DFDMA

0

1

1

1

2

0.95

1

0.96

4

0.002

0.99

0.0023

6

0.000001

0.2

0.000001

8

0.000002

0.002

0.0000023

10

0.0000025

0.000001

0.0000026

PAPR and SER Performance Analysis of OFDMA and SCFDMA

139

mapping. (right side figure is added as a subplot to show the values as they are overlapped in left one) Conclusion: In this project we have focused on the performance analysis of OFDMA and SCFDMA systems in terms of SER and PAPR. Three types of channel (pedAchannel, vehAchannel and AWGN channel) were introduced along with two types of subcarrier mapping (localised and distributed, IFDMA is a special case of DFDMA). It was concluded that the PAPR performance of SCFDMA was significantly better in all three subcarrier mapping techniques i.e LFDMA, IFDMA and DFDMA, when compared to their counterparts in OFDMA.

References 1. Matousek, R., Osmera, P., & Roupec, J. (2000). GA with fuzzy inference system. In Proceedings of the 2000 Congress on Evolutionary Computation, La Jolla, CA, USA (Vol. 1, pp. 646–651). Retrieved July 16–19, 2000; CEC00 (Cat. No. 00TH8512). 2. Krishna, S. L. V., Abdul Rahim, B., Shaik, F., & Rajan, K. S. (2010). Lossless embedding using pixel differences and histogram shifting technique. In Recent Advances in Space Technology Services and Climate Change 2010 (RSTS & CC-2010) (pp. 213–216). Chennai, India. https:// doi.org/10.1109/RSTSCC.2010.5712850 3. Siegel, J. E., Erb, D. C., & Sarma, S. E. (2018). A survey of the connected vehicle landscape—Architectures, enabling technologies, applications, and development areas. In IEEE Transactions on Intelligent Transportation Systems (TITS) (Vol. 19, No. 8). 4. Lakshmanna, K., Shaik, F., Gunjan, V. K., Singh, N., Kumar, G., & Shafi, R. M. (2022). Perimeter degree technique for the reduction of routing congestion during placement in physical design of VLSI circuits. Complexity, 2022, Article ID 8658770, 11 p. https://doi.org/10.1155/ 2022/8658770 5. Akhtar, M., Hannan, M., Begum, R., Basri, H., & Scavino, E. (2017). Backtracking search algorithm in CVRP models for efficient solid waste collection and route optimization. Waste Management, 61, 117–128. 6. Hindy, A., Mittal, U., & Brown, T. (2020). CSI feedback overhead reduction for 5G massive MIMO systems. In 2020 10th Annual Computing and Communication Workshop and Conference (CCWC) (pp. 0116–0120). IEEE. 7. Surya Narayana, G., Kolli, K., Ansari, M. D., & Gunjan, V. K. (2021). A traditional analysis for efficient data mining with integrated association mining into regression techniques. In ICCCE 2020: Proceedings of the 3rd International Conference on Communications and Cyber Physical Engineering (pp. 1393–1404). Springer Singapore. 8. Baeza, V. M., & Armada, A. G. (2019). Noncoherent massive MIMO. In Wiley 5G ref: The essential 5G reference online (pp. 1–28). 9. Nagaraju, C. H., & Kondagandla, R. (2022). IoT based live monitoring public transportation security system by using raspberry Pi, GSM& GPS. In V. K. Gunjan, & J. M. Zurada (Eds.), Modern approaches in machine learning & cognitive science: A walkthrough. Studies in computational intelligence (Vol. 1027). Springer, Cham. https://doi.org/10.1007/978-3-03096634-8_43 10. Karimullah, S., Vishnuvardhan, D., & Bhaskar, V. (2022). An improved harmony search approach for block placement for VLSI design automation. Wireless Personnel Communications, 127, 3041–3059. https://doi.org/10.1007/s11277-022-09909-2 11. Jaya Krishna, N., Shaik, F., Harish Kumar, G. C. V., Naveen Kumar Reddy, D., & Obulesu, M. B. (2021). Retinal vessel tracking using Gaussian and radon methods. In A. Kumar, & S. Mozar

140

12.

13. 14.

15. 16.

17.

18.

19.

20. 21.

G. Obulesu et al. (Eds.), ICCCE 2020. Lecture notes in electrical engineering (Vol. 698).Singapore: Springer. https://doi.org/10.1007/978-981-15-7961-5_37 Kahng, A. B. (2018). INVITED: Reducing time and effort in IC implementation: A roadmap of challenges and solutions. In Proceedings of the Proceedings of the 55th Annual Design Automation Conference, San Francisco, CA, USA (pp. 1–6). Retrieved June 24–29, 2018 Wang, C., Qin, J., Qu, C., Ran, X., Liu, C., & Chen, B (2021). A smart municipal waste management system based on deep-learning and Internet of Things. Waste Manag. 135, 20–29. Karimullah, S., Vishnu Vardhan, D., & Basha, S. J. (2020). Floorplanning for placement of modules in VLSI physical design using harmony search technique. In ICDSMLA 2019. Lecture notes in electrical engineering (Vol. 601). Springer Nature Singapore Pte Ltd. Kansal, P., & Shankhwar, A. K. (2017). FBMC versus OFDM waveform contenders for 5G wireless communication system. Wireless Engineering and Technology, 8(4), 59–70. Karimullah, S., &Vishnu Vardhan, D. (2020). Iterative analysis of optimization algorithms for placement and routing in Asic design. In ICDSMLA 2019. Lecture notes in electrical engineering (Vol. 601). Springer Nature Singapore Pte Ltd. Madhuravani, B., Krishnasrija, R., & Degala, D. P. (2022). Early prediction of chronic kidney disease using predictive analytics. In Confidential computing: Hardware based memory protection (pp. 39–46). Singapore: Springer Nature Singapore. Hemalatha, V., Premalatha, B., & Kiran Kumar, K. (2022). Dual security based attendence system by using face recognition and RFID with GSM. In Confidential computing: Hardware based memory protection (pp. 9–18). Singapore: Springer Nature Singapore. Singh, N., Gunjan, V. K., & Nasralla, M. M. (2022). A parametrized comparative analysis of performance between proposed adaptive and personalized tutoring system “Seis Tutor” with existing online tutoring system. IEEE Access, 10, 39376–39386. https://doi.org/10.1109/ACC ESS.2022.3166261 Chadchan, S., & Akki, C. B. (2010). 3GPP LTE/SAE: An overview. International Journal of Computer Science and Engineering Technology (IJCEE), 2(2), 806–814. Surya Narayana, G., Kolli, K., Ansari, M. D., & Gunjan, V. K. (2021). A traditional analysis for efficient data mining with integrated association mining into regression techniques. In A. Kumar, & S. Mozar (Eds.), ICCCE 2020. Lecture notes in electrical engineering (Vol. 698). Singapore: Springer. https://doi.org/10.1007/978-981-15-7961-5_127

Food Detection with Image Processing Using Convolutional Neural Network (CNN) K. Sreenivasa Rao, Fahimuddin Shaik, Munaga Sai Krishna, Sompalli Bhavya, Pothalam Bharat Teja, and Saginala Jaleel Basha

1 Introduction In terms of meeting client needs, the restaurant is one of the most important aspects of the food service sector as well as a part of tourism. The restaurant is organised into several sections, one of which is the canteen [1]. A canteen is a cafeteria found in workplaces, factories, and schools where employees and students can eat lunch, take a break with refreshments, or study while on the job [2]. The influx of customers at various times of the day has caused a backlog in food payments. A line may form if the facility’s capacity to provide the service exceeds the quantity of demand. Queues are common in restaurants during lunch and dinner [3]. Due to the vast array of food varieties, image recognition of food products is frequently difficult [4]. The goal of food detection is to make paying at restaurants easier, and automatic food price estimation is accomplished through the use of the network [5]. We applied AlexNet to food identification and recognition tasks using image processing techniques. We created a dataset of the most common food products in kaggle and used it to test recognition skills. The network categorises the train and test data and outputs the classification results.

K. S. Rao · F. Shaik (B) · M. S. Krishna · S. Bhavya · P. B. Teja · S. J. Basha Electronics and Communication Engineering, Annamacharya Institute of Technology and Sciences, Rajampet, India e-mail: [email protected] K. S. Rao e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 V. K. Gunjan et al. (eds.), Modern Approaches in Machine Learning and Cognitive Science: A Walkthrough, Studies in Computational Intelligence 1117, https://doi.org/10.1007/978-3-031-43009-1_13

141

142

K. S. Rao et al.

2 Literature Review Modern food identification applications depend heavily on emerging techniques to food categorization. For this reason, a new food recognition algorithm is offered that considers the shape, colour, size, and texture of the object [6]. A more accurate classification can be achieved by combining these criteria in different ways. Segmentation and classification are this application’s two primary procedures. Segmentation is the technique of breaking a digital image up into various pieces in image processing [7]. By lowering an image’s complexity, segmentation aims to increase its meaning. The major objective of this strategy is to use a network to identify food and determine total costs. The data is trained using the AlexNet algorithm [8]. Using image processing methods, labels are applied to the input image, and a bounding box is created. The network makes a distinction between the test image and the train dataset. In order to clarify the detection processes utilised for identifying the underlying concerns and emphasise the preventative steps to solve the problem, a comparative study of food intake detection using artificial neural networks and support vector machines was carried out in [9]. The Real-Time Mobile Food Recognition System, which is detailed in [10], explains the important aspects of the numerous issues that must be resolved in the Realtime environment.

3 Methodology 3.1 Existing Method Finding objects and boundaries (lines, curves, etc.) in photographs is a common usage of image segmentation [11, 12]. Pixels in the same area share characteristics like colour or texture. In this application, foods with similar constituents will be grouped together in the same segment. Therefore, dividing the meal into distinct portions helps the classifier get the right answer [13, 14]. Taking an image and performing segmentation may seem simple, but a strong model is necessary to achieve outstanding results during the segmentation and classification phases [15].

3.2 Proposed Method This study employs a technique to enhance the identification of by training a wider sample of images and calculating the total cost of the meal. The dataset is trained using AlexNet, which also carries out classification. The block diagram for the suggested

Food Detection with Image Processing Using Convolutional Neural …

143

method, which is depicted in Fig. 1, describes how the work is done based on training and testing methods. The process of transforming a grayscale image into a bi-level image is known as image binarization. Black and white pixels are the two categories into which pixels in an image are separated. The primary goal of picture binarization is to divide an image into the foreground and background. In essence, it turns the image’s 256 shades of grey into just two—black and white—creating a binary image. In response to the set of qualities specified by attributes, it delivers measurements for each of the eight connected components (objects) in the binary picture, BW. Each photo object is represented by a Struct in the Stats structure array. Both contiguous and discontiguous territories can employ region props. Alex Net: Using a convolutional neural network, Alex Net classified a fresh set of pictures. Alex Net can categorise photos into over a thousand different object categories after training on over a million photographs as shown in Fig. 2. The network includes a wealth of features for a wide variety of pictures. The image layer input of a convolutional neural network specifies the input pixels and comprises of the raw input values of the images. Utilize the image Input function to create a layer. In the input Size argument, specify the hue of your image. The first layer to extract attributes from an input image is convolution. The inputs needed are an image matrix and a filter or kernel. The output matrix is presented in Fig. 3 and the matrix multiplication is depicted in Fig. 4. A convolutional layer’s neurons were connected to the layer preceding it. A convolutional layer learns the information localised by these regions while scanning over a picture. The pixels of these zones can be specified using the filter Size input argument in the call to the convolution2dLayer function. One can additionally add zero to the vertical and horizontal picture output by using the ‘Padding’ name-value pair option. Padding involves extending the borders of the image input by zeros in either rows or columns. It supports modifying the layer’s output. This value must be an integer in order for the complete image to be covered. The software will by default ignore any remaining portions of the image if the combination of these parameters is unable to completely cover them.

4 Conclusion Using deep learning techniques and visual processing, the suggested method may identify the food. The network is trained using the Kaggle dataset. Existing technologies can process, assess, and recognise fruits and meals based on colour and texture cues. We were able to increase the adaptability and capabilities of the recognition system using AlexNet. A significant amount of data augmentation is required, in addition to learning the generalised pattern required to identify and recognise food

Fig. 1 Proposed method process flow

144 K. S. Rao et al.

Food Detection with Image Processing Using Convolutional Neural …

145

Fig. 2 Steps in the classification of Image using AlexNet Fig. 3 Output matrix

Fig. 4 Multiplication of Image matrices

items. The network separates the data into train data and test data categories. The accuracy is far higher than any other conventional models.

146

K. S. Rao et al.

References 1. Kahng, A. B. (2018). INVITED: Reducing time and effort in IC implementation: A roadmap of challenges and solutions. In Proceedings of the Proceedings of the 55th Annual Design Automation Conference, San Francisco, CA, USA. (pp. 1–6). Retrieved June 24–29, 2018. 2. Gunjan, V. K., Kumar, S., Ansari, M. D., & Vijayalata, Y. (2022). Prediction of agriculture yields using machine learning algorithms. In Proceedings of the 2nd International Conference on Recent Trends in Machine Learning, IoT, Smart Cities and Applications: ICMISC 2021 (pp. 17– 26). Springer Singapore. 3. Sharma, S., & Rai, B. K. (2023). Biomedical data classification using fuzzy clustering. In AI and blockchain in healthcare (pp. 83–92). Singapore: Springer Nature Singapore. 4. Rudra Kumar, M., Pathak, R., & Gunjan, V. K. (2022). Machine learning-based project resource allocation fitment analysis system (ML-PRAFS). In Computational Intelligence in Machine Learning: Select Proceedings of ICCIML 2021 (pp. 1–14). Singapore: Springer Nature Singapore. 5. Kaur, I., Gupta, V., Verma, V., & Kaur, S. (2023). Securing healthcare records using blockchain: applications and challenges. In AI and blockchain in healthcare (pp. 57–66). 6. Angeline, R., Kanna, S. N., Menon, N. G., & Ashwath, B. (2022). Identifying malignancy of lung cancer using deep learning concepts. Artificial intelligence in healthcare, 35–46. 7. Jaya Krishna, N., Shaik, F., Harish Kumar, G. C. V., Naveen Kumar Reddy, D., & Obulesu, M. B. (2021). Retinal vessel tracking using Gaussian and radon methods. In A. Kumar, & S. Mozar (Eds.), ICCCE 2020. Lecture notes in electrical engineering (Vol. 698). Singapore: Springer. https://doi.org/10.1007/978-981-15-7961-5_37 8. Kumar, S., Ansari, M. D., Gunjan, V. K., & Solanki, V. K. (2020). On classification of BMD images using machine learning (ANN) algorithm. In ICDSMLA 2019: Proceedings of the 1st International Conference on Data Science, Machine Learning and Applications (pp. 1590– 1599). Singapore: Springer. 9. Guo, Z., Mai, J., & Lin, Y. (2021). Ultrafast CPU/GPU kernels for density accumulation in placement. In Proceedings of the 2021 58th ACM/IEEE Design Automation Conference (DAC), San Francisco, CA, USA (pp. 1123–1128). Retrieved December 5–9, 2021. 10. Patibandla, R. L., Srinivas, V. S., Rao, B. T., & Murthy, M. R. (2021). Pneumonia prediction using swarm intelligence algorithms. In Artificial intelligence in healthcare (pp. 103–116). Singapore: Springer Singapore. 11. Handoko, W. M. W., & Widjojo, A. R. (2013). Analisis tingkat pelayanan optimal pdarumahmakan mie ayam mas yudi jl. sagan kidul no. 20 yogyakarta. Journal Universitas Atma Jaya Yogyakarta, 25, 73–89. 12. Pan, J., Chang, C.C., Xie, Z., Li, A., Tang, M., Zhang, T., Hu, J., & Chen, Y. (2022). Towards collaborative intelligence: Routability estimation based on decentralized private data. In Proceedings of the DAC ’22, 59th ACM/IEEE Design Automation Conference, San Francisco, CA, USA (pp. 961–966). Retrieved July 10–14, 2022. 13. Ristiawanto, V., Irawan, B., & Setianingsih, C. (2019). Wood classification with transfer learning method and bottleneck features. In 2019 International Conference on Information and Communications Technology (ICOIACT). 14. Surya Narayana, G., Kolli, K., Ansari, M. D., & Gunjan, V. K. (2021). A traditional analysis for efficient data mining with integrated association mining into regression techniques. In ICCCE 2020: Proceedings of the 3rd International Conference on Communications and Cyber Physical Engineering (pp. 1393–1404). Singapore: Springer. 15. Eduardo, A., Beatriz, R., Marc, B., & Petia, R. (2020) Grab pay and eat: Sematic food detection for smart restaurants. IEEE Transactions on Multimedia, 1(1).

Google Appstore Data Classification Using ML Based Naïve’s Bayes Algorithm: A Review J. Chinna Babu, Y. Suresh, Ajmeera Kiran, A. Ramesh Babu, and C. Madana Kumar Reddy

1 Introduction Mobile phones have recently shown to be helpful in our daily lives, especially with the increased availability and decreased cost. In many fields, machine learning has been used extensively to identify patterns in data and learn from them in order to make precise predictions. In our daily lives, machine learning is used frequently, particularly by email providers to identify spam messages. Due to the vast amount of data that must be analyzed, understanding and reaching the right decisions can be difficult as a result of the recent increase in data. Making accurate decisions can also be difficult because the data is available in so many different formats. The company’s primary app store, Google Play Store, is used daily by millions of users. It allows J. C. Babu (B) Department of Electronics and Communication Engineering, Annamacharya Institute of Technology and Sciences, Rajampet 516126, Andhra Pradesh, India e-mail: [email protected] Y. Suresh Department of Information Technology, PVP Siddhartha Institute of Technology, Vijayawada, India A. Kiran Department of Computer Science and Engineering, MLR Institute of Technology, Hyderabad, Telangana 500043, India e-mail: [email protected] A. R. Babu Department of Artificial Intelligence and Data Science, Annamacharya Institute of Technology and Sciences, Rajampet, India C. M. K. Reddy Head of the Department, Department of Computer Applications, Annamacharya Institute of Technology and Sciences, Rajampet 516126, Andhra Pradesh, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 V. K. Gunjan et al. (eds.), Modern Approaches in Machine Learning and Cognitive Science: A Walkthrough, Studies in Computational Intelligence 1117, https://doi.org/10.1007/978-3-031-43009-1_14

147

148

J. C. Babu et al.

you to access a variety of stuff, including audio recordings, video games, watching television, and publications. Users may rate an app after downloaded it from the app stores to share their individual opinions with other people using it. This is reciprocal since other users’ evaluations may inspire them.

2 Literature Review 2.1 Random Forest Algorithm Based on Data Classification—Machine Learning Random forest is a renowned computational learning framework that makes use of the supervised training method. It may be used to solve categorization and regressionrelated ML problems [1]. According to the concept of collaborative learning, which is the integration of several classifiers to solve a complex problem and improve the efficiency of an algorithm [2], as shown in Fig. 1, it relies on the situation. Results: After merging the outcomes of all test models, the final decision is made based on a majority vote [3]. Aggression is the process of collecting all the findings and producing output based on the majority opinion [4]. So this can handle binary, continuous, and categorical data. The output results can be obtained by importing the libraries, importing the data sets and putting feature variables to target variables,

Fig. 1 Pictorial representation of data classification using Random Forest algorithm

Google Appstore Data Classification Using ML Based Naïve’s Bayes …

149

Fig. 2 The visualization produced by the RF classifier when using the data

train-test-split is performed and then random forest classifier is imported and fitted into the data. Now visualization of the sorted data is performed [5]. Machine Learning’s role in classification of data: Limitations: While random forest may be utilized for both regression and classification methods which was demonstrated in Fig. 2, it is not better suited for Regression activities.

2.2 Classification of Data Using Decision Tree Classifier-Machine Learning Techniques Decision Tree Classifier is a popular machine learning technique used for classification tasks. It creates a tree-like model of decisions and their possible consequences, which helps in classifying new data points based on their features. Here’s a step-by-step guide on how to perform data classification using a Decision Tree Classifier: Data Preparation: Start by preparing your dataset for classification. Ensure that your data is in a structured format, with rows representing individual samples and columns representing features. Additionally, make sure that your dataset is labeled, meaning that each sample has a corresponding class or category assigned to it. Feature Selection: Choose the pertinent characteristics from your dataset that will be utilized for categorization in the feature choice step. The issue domain and the type

150

J. C. Babu et al.

of data both influence the feature selection. The precision of the categorization model depends heavily on the choice of informational and discriminatory characteristics. Splitting the Dataset: The set used for training and the testing set should be separated from your dataset. The decision tree classifier will be constructed using the set for training, and its effectiveness will be assessed using the set for testing. Usually, between 70 and 80% of the data are in the training set, while between 20 and 30% are in the set used for testing. Building the Decision Tree: Create a decision tree classifier using the training set as the basis. The decision tree method divides the data recursively depending on the chosen characteristics to produce a tree-like structure. The most popular approach for creating decision trees is the CART (Classification and Regression Trees) algorithm, which chooses the optimal split at each node using a metric like impurity from Gini or entropy. Training the Decision Tree: Train the decision tree classifier using the training set. The algorithm learns the patterns and relationships between the features and the corresponding classes. It iteratively splits the data based on feature values to maximize the purity of each resulting branch. Evaluating the Decision Tree: Once the decision tree is trained, evaluate its performance using the testing set. Pass the testing set through the decision tree, and compare the predicted classes with the actual classes in the dataset. Common evaluation metrics for classification models include accuracy, precision, recall, and F1-score. Tuning and Pruning: By modifying hyper parameters like the maximum depth of the tree and the bare minimum number of information needed to divide a node, you may fine-tune the decision tree classifier. Pruning, a method used to prevent overfitting, which enhances the model’s capacity for generalization. Predicting New Data: Finally, you can use the trained decision tree classifier to predict the classes of new, unseen data points. Simply pass the new data through the decision tree, following the decision rules at each node, until you reach a leaf node that corresponds to the predicted class. That’s the basic workflow for performing data classification using a Decision Tree Classifier. Remember that Decision Trees have both advantages and limitations, and the choice of the appropriate algorithm depends on the specific problem and the characteristics of the data. Regenerate response shown in Fig. 3. In order to build a tree, this paper used the CART algorithm. A decision tree simply asks a question, and based on the answer (Yes/No), it further split the tree into subtrees [6]. Results: Each time this method iterates, the output results are dependent on the characteristics’ entropy (H) and information gain (IG). To the extent that the halting requirements are not met, the splitting process produces fully formed trees. In order to overcome the attribute selection issue, the Gini index, Gain ratio, Reduction in variance, and

Google Appstore Data Classification Using ML Based Naïve’s Bayes …

151

Fig. 3 Block diagram of machine learning based decision tree classifier

Chi-square will be calculated here. This method keeps iterating over each subset while only taking into account traits that have never been chosen previously. As can be seen in Figs. 4 and 5, the pruned model is therefore simpler, more comprehensible, and easier to grasp than the random forest model. Limitations: a. The process of choosing a tree is complicated since it has several tiers. b. It might have a problem with excessive overfitting. c. When relative to other methods, less information processing is necessary.

3 Classification of Data Using Support Vector Machines-Machine Learning Support Vector Machines are categorized into more specialized computer science subfields, and text mining may transform a shape into an organized form using homogenous data [7, 8]. Support Vector Machine (SVM), one of the supervised machine learning techniques, is used in the above classification in order to speed up the entire procedure compared to the human approach [9]. In order to model abstract words with SVM, the TF-IDF approach is utilized to turn them into vectors. The 10-fold cross validation method will be used with the classification model. Based on these categorizations the matrix of confusion computation and the application of

152

J. C. Babu et al.

Fig. 4 The above image is the visualization result for the decision tree classifier working with the test set result Fig. 5 RF algorithm working with the training data set result

three SVM kernels will be used to determine the performance that will result [10]. Based on the findings of this study, it can be concluded that there are two variables that influence categorization precision: the quantity of participants in scientific classes that are unbalanced and the quantity of characteristics derived through the use of text mining [11]. The SVM Linear Kernel and 205 characteristics were used to achieve the highest degree of precision of examination, yielding a value of 58.3% (Fig. 6). Results: As illustrated in Fig. 7, the resulting outcomes are generated by using the hyper plane in an N-dimensional space known as Euclidean, which is a flat, the subset dimension is N−1of that space and separates it into two unconnected routes. Thus, by introducing

Google Appstore Data Classification Using ML Based Naïve’s Bayes …

153

Fig. 6 Generic block diagram of data classifier using support vector machines

a new dimension to make the data linearly distinct and then applying algebraic manipulations to handle linear and non-linear issues, this paper may categorize the data by reflecting the boundary of decision back to the initial measurements.

Fig. 7 The result of test dataset

154

J. C. Babu et al.

Limitations: 1. It is hard to choose an appropriate Kernel form: For handling the non-linear information, it is tricky to choose the optimum Kernel functions. It could be difficult and complicated. You run the danger of creating too many vectors of support when using a big size kernel, which significantly slows down training. 2. The SVM method requires a significant amount of memory. You need a lot of storage since you have to retain all the support vectors in memory and because the size of the training dataset pushes this amount up fast. 3. Requires Feature Sizing: Feature scaling of the variables is necessary beforehand to employing SVM. 4. SVM training on big data sets takes quite a while. 5. Difficult to comprehend: SVM model used is challenging.

4 Classification Using Logistic Regression Algorithm-Machine Learning One of the first and most used Machine Learning computations, logistic regression (LR) belongs to the category of supervised training [12]. Using a predetermined set of independent variables, it is used to forecast a related categorized variable. As seen in Fig. 8, a model of logistic regression predicts the results of a variable with a category that is dependent. The outcome must thus be a discrete or classification value. The exact value might be stated as 0 or 1, but it could also be described as either Yes or No, 0 or 1, True or False, etc. It gives probabilities in the range of 0 to 1 values. It is quite similar to the application of linear regression, with the possible exception of how they are employed [13]. For classification problems, a Logistic Approach to Regression is employed; for regression problems, a linear-regression technique is employed [14].

Fig. 8 The simulation of threshold value test set result

Google Appstore Data Classification Using ML Based Naïve’s Bayes …

155

Fig. 9 Prediction results of linear and logistic prediction

Results: Here the output results can be obtained by using confusion matrix and ROC curves, maximum likelihood and sigmoid function. For logistic regression outcomes are target variables are dichotomous in nature. In order to forecast the output, this method of statistics for predicting binary classes computes the likelihood of an occurrence. In contrast, the goal variable in linear regression is a qualitative one. Therefore, logistic regression produces an unchanged result as illustrated in Fig 9 and employs a log of chances as the variable that is dependent. Limitations: 1. In real life, data is rarely linearly separable, hence the presumption of linearity between the dependent and independent variables is a major drawback of the logistic regression method. The majority of the time, information will appear disorganized. 2. Regression using logistic regression should not be employed if there are less data than features, since this might result in overfitting. 3. Discrete functions can only be predicted using logistic regression methods. Logistic regression can only use the discrete number set as the dependent variable. Due to the excessively constant information, this constraint in and of itself is difficult. 4. Logistic regression models can be outperformed by more advanced algorithms like artificial neural networks and randomly generated forests. 5. The technique for logistic regression is dependent on exceptions.

156

J. C. Babu et al.

5 Data Classification Using KNN Algorithm-Machine Learning K-Nearest Neighbours represents one of the simplest supervised learning-based autonomous training algorithms. The KNN approach assigns new data to the category that is most similar to existing categories under the presumption that new data and current events are related. The procedure known as KNN categorizes new data points based on similarity and records all readily available information [15]. This suggests that when new data is produced, it may be swiftly and precisely categorized using the technique known as KNN. As shown in Fig. 10, the KNN approach may be applied to classification and regression problems, however it is more typically employed for issues related to classification. KNN is a non-parametric in method, which means it doesn’t make any presumptions about the data it uses [16]. Since it stores a dataset instead of immediately learning from the training set, it is also known as a lazy learner algorithm [17, 18]. Instead, it groups recent data into a category that closely resembles the recent information. Results: The degree of similarity among the fresh data points and the previously stored data points is used to determine the output results. Parameter tuning, which is required for improved outcomes, is the process of selecting the appropriate value of K [19]. By computing the Euclidean distance between the newly inserted data point and its closest neighbours, the outcome may be predicted. Our goal is to find the shortest Euclidean distance, and the base of estimation depends on how many lesser distances are there [20].

Fig. 10 New data element added into existing data and added data is classified using KNN algorithm

Google Appstore Data Classification Using ML Based Naïve’s Bayes …

157

6 Limitations 1. It is constantly necessary to find the value of K, which may be at times difficult. 2. Because the procedure must establish the distance among every information point in every instance of training, it is expensive. Conclusion: The Random Forest Algorithm, Tree Classifier based on Decision, Regression based on Logistics, Support Vector Machines, and KNN algorithm are only a few of the approaches for data categorization based on ML techniques that are described in this Paper. We classified the provided data using several methods, and by selecting various techniques, we will obtain various outcomes. While Naive’s Bayes is a classification technique for situations involving binary and many classes. When this method is explained using binary or categorical input values, it is easier to grasp. The mainstream of requests for this Naive Bayes method including analysis of sentiment, filtering spam and the systems recommendations, etc. This is simple and fast to put into practice, but the predictors’ dependence in most real-world scenarios makes the classifier less effective.

References 1. Zhou, P., Yang, X. L., Wang, X. G., & Hu, B., et al. (2020). A pneumonia outbreak associated with a new coronavirus of probable bat origin. Nature. 2. Gunjan, V. K., Prasad, P. S., Fahimuddin, S., & Bigul, S. D. (2019). Experimental investigation to analyze cognitive impairment in diabetes mellitus. In: A. Kumar, S. Mozar (Eds.), ICCCE 2018. Lecture notes in electrical engineering (Vol. 500). Singapore: Springer. https://doi.org/ 10.1007/978-981-13-0212-1_79 3. Babu, J. C., Kumar, M. S., Jayagopal, P., Sathishkumar, V. E., Rajendran, S., Kumar, S., Karthick, A., & Mahseena, A. M. (2022). IoT-based intelligent system for internal crack detection in building blocks. Journal of Nanomaterials, 2022, Article ID 3947760, 8. 4. Rudra Kumar, M., Pathak, R., & Gunjan, V. K. (2022). Machine learning-based project resource allocation fitment analysis system (ML-PRAFS). In Computational Intelligence in Machine Learning: Select Proceedings of ICCIML 2021 (pp. 1–14). Singapore: Springer Nature Singapore. 5. Chinna Babu, J., & Shankar, K. (2022). Melanoma skin segmentation process using PCA and morphological methods. In: Modern approaches in machine learning & cognitive science: A walkthrough-studies in computational intelligence-Springer (Vol. 1027). 6. Chinna Babu, J., & Naveen Kumar Raju, K. (2022). Safety locker system with image identification by using IOT. In Modern approaches in machine learning & cognitive science: A walkthrough (Vol. 1027, pp. 415–422). Studies in Computational Intelligence-Springer. 7. Venkatesh, B., Babu, J. C., Mathivanan, S. K., Jayagopal, P., Prasanna, S., & Uddin, M. S. (2022). Influences of aqueous nanofluid emulsion on diesel engine performance, combustion, and emission: IoT (Emission monitoring system). Advances in Materials Science and Engineering, 2022, Article ID 8470743. 8. Hussain, S. A., Babu, J. C., Hasan, R., & Mahmood, S. (2022). A hybrid soft bit flipping decoder algorithm for effective signal transmission and reception. TELKOMNIKA (Telecommunication Computing Electronics and Control), 20(3), 510–518.

158

J. C. Babu et al.

9. Gill, H. S., Khalaf, O. I., Alotaibi, Y., Alghamdi, S., & Alassery, F. (2022). Fruit image classification using deep learning. CMC-Computers Materials & Continua, 5135–5150. 10. Babu, J. C., Kumar, M. S., Jayagopal, P., Sathishkumar, V. E., Rajendran, S., Kumar, S., Karthick, A., & Mahseena, A. M. (2022).IoT-based intelligent system for internal crack detection in building blocks. Journal of Nanomaterials, 1–14. 11. Janniekode, U. M., Somineni, R. P., Khalaf, O. I., Itani, M. M., Chinna Babu, J., Abdulsahib, G. M. (2022). A symmetric novel 8T3R non-volatile SRAM cell for embedded applications. Symmetry, 768–772. 12. Jhun, M., & Huh, M. -H. (2001). Random permutation testing in multiple linear regression. Communications in Statistics-Theory and Methods, 2023–2032. 13. Ogudo, K. A., Surendran, R., & Khalaf, O. I. (2023). Optimal artificial intelligence based automated skin lesion detection and classification model. Computer Systems Science and Engineering, 693–707. 14. Dave, K., Lawrence, S., & Pennock, D. M. (2003). Mining the peanut gallery: Opinion extraction and semantic classification of product reviews. In World Wide Web (pp. 59–528). 15. Dave, K., Lawrence, S., & Pennock, D. M. (2003). Mining the peanut gallery: Opinion extraction and semantic classification of product reviews. In World Wide Web (pp. 519–528). New York, USA. 16. Gunjan, V. K., Vijayalata, Y., Valli, S., Kumar, S., Mohamed, M. O., & Saravanan, V. (2022). Machine learning and cloud-based knowledge graphs to recognize suicidal mental tendencies. Computational Intelligence and Neuroscience. 17. Khaparde, A. R., Alassery, F., Kumar, A., Alotaibi, Y., & Khalaf, O. I. (2022). Differential evolution algorithm with hierarchical fair competition model. Intelligent Automation & Soft Computing, 1045–1062. 18. Ting, L., Ip, W. H., Tsang, A. H. C. (2011). Is Naïve bayes a good classifier for document classification. International Journal of Software Engineering and its Applications, 5(3), 37–46. 19. Pagano, D., & Maalej, W. (2013). User feedback in the app store: an empirical study. In IEEE (pp. 125–134). 20. Liu, Y. (2022). Interaction-enhanced and time-aware graph convolutional network for successive point-of-interest recommendation in travelling enterprises. IEEE Transactions on Industrial Informatics, 10.

Improved Spectral Efficiency Using Vehicular Visible Light Communication with 16-Bit DCO in OFDM Shaik Karimullah, D. Vishnuvardhan, Vinit Kumar Gunjan, and Fahimuddin Shaik

1 Introduction Next-generation transportation trends that involve self-driving cars and ride sharing need the use of various vehicular networking systems [1]. In contrast, intelligent modes of transportation (ITS) that are connected with vehicular communications aim to reduce congestion in traffic, crashes, air pollution, consumption of energy, and time waste [2]. Up-to-date vehicle communications are intended to enable timely and efficient delivery of information regarding accidents, congestion in traffic, and roadway conditions that drivers do not comprehend [3]. Furthermore, OFDM with QAM is highly suited for broadband applications due to its capacity to handle frequencyselective channels [4]. QAM offers resistance to fading and interference on each subcarrier, while OFDM allows each subcarrier to experience a variety of channel conditions separately [5]. As a result, OFDM with QAM is suitable for systems like wireless communication, digital television, and broadband internet access [6]. In broadband applications, combining OFDM with QAM offers a strong method for delivering high data rates, efficient spectrum utilisation, and dependable transmission in challenging channel conditions [7, 8]. It has developed into the foundation of modern broadband communication networks, offering seamless connectivity, highspeed internet access, and delivery of multimedia content in a variety of contexts. S. Karimullah (B) · F. Shaik Department of ECE, AITS, Rajampet, India e-mail: [email protected] D. Vishnuvardhan Department of ECE, JNTUACE, Ananthapuramu, India e-mail: [email protected] V. K. Gunjan Department of Computer Science and Engineering, CMR Institute of Technology Hyderabad, Hyderabad, India © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 V. K. Gunjan et al. (eds.), Modern Approaches in Machine Learning and Cognitive Science: A Walkthrough, Studies in Computational Intelligence 1117, https://doi.org/10.1007/978-3-031-43009-1_15

159

160

S. Karimullah et al.

[9]. VLC systems provide substantial benefits over conventional systems [10] in following areas. 1. 2. 3. 4.

High Bandspectrum. More secure and easy to implement. Highly immune to interference with electromagneticsignals. Robust againstRoping.

These essential qualities of the VLC system enable it to be employed in high-fidelity systems that support the web and the IOT-Internet of Things, and also in medical health-care systems to provide data service as well as tracking in areas when radio frequency has constraints [11, 12]. With the use of traffic light infrastructure, VLC may also be employed in car devices for lane shift warning, traffic sign motion assistant, pre-crash detection, and other purposes [13].

2 Literature Review Since the data that is broadcast is comparable on a number of different rates, the bit sequence time is significantly longer compared to a sequential system having the same total data rate. Because a symbol’s time is longer, ISI only affects one symbol, making equalisation easier. Any residual ISI in maximal OFDM operations is indifferent, spending a cyclic prefix, resulting in a sort of guard interval [14]. When Frequency Division Multiplexing (FDM) is used in unadventurous wireless networks or Wavelength Division Multiplexing (WDM) is used in photosensitive schemes, data is delivered on a number of separate regularities concurrently. However, there are some key conceptual and real-world contrasts between OFDM and these classical systems. The subchannel frequency were chosen across one OFDM symbols period such that the indications are scientifically false. The required orthogonal signs may be generated precisely and mathematically efficiently by using an Inverse Fast Fourier Transform (IFFT) for both modulation and multiplexing. In FDM or WDM, frequency difference bands are located around the subcarriers. Individual subchannels are recovered at the receiver using analogue filtering methods. Since the area of each OFDM subcarrier takes on a form, each OFDM subchannel contains enormous side portions throughout a frequency range that includes numerous other subchannels. One of the primary disadvantages of OFDM is its susceptibility to frequency imbalance and phase clatter [15]. By successfully tackling the issues of multi-path fading, intersymbol interference, and frequency-selective fading, Orthogonal Frequency Division Multiplexing (OFDM), a modulation technology, has revolutionised communication networks. Due to its special qualities and advantages, OFDM, which was first created for wireless applications, has also showed promise in optical display systems. In this review, we will discuss the fundamentals of OFDM that are important for optical display systems and dispel several myths that have previously circulated among

Improved Spectral Efficiency Using Vehicular Visible Light …

161

OFDM researchers. These subcarriers together cover the full available bandwidth, with each subcarrier often being narrowband and occupying a certain frequency range. For applications that have high bandwidth needs, such optical displays, this parallel data transfer provides high data rates and effective utilisation of the available spectrum. The subcarriers’ orthogonality is a crucial aspect of OFDM. Even when they are close together, it makes sure that the subcarriers don’t interfere with one another. The subcarriers’ frequencies and phases are carefully designed in order to achieve this orthogonality. As a result, the subcarriers may be quickly and individually demodulated and decoded, increasing the transmission’s resilience and dependability. The ability of OFDM to split the accessible frequency spectrum into several orthogonal subcarriers is one of its key features. Constructed on real vehicle To Vehicle (V2V) transmission data, the normalized Channel Frequency Response for the V2LC network established to be invariable of inter-vehicle distance, TX/RX ambient light and also zenith angle. This medium characteristic is then employed in the BCE to approximate only the normalization factor, not the CSI, with all subcarriers independently. The projected method outclasses pilot-based CE methods on the basis of throughput performance and the bit error rate (BER) for any and all modulation techniques, with the exception of the 64-QAM DCO- OFDM, that is not practical because there is no suitable estimator, according to widespread simulations at dissimilar speeds of vehicle. Additionally, Simulation of Urban Mobility’s realistic vehicle mobility scenario shows that the performance in real-time of the suggested BCE is very close to each modulation scheme’s maximum throughput at high signal-to-noise ratio (SNR) levels (SUMO). Advantages: • High speed communication • Accurate Channel Estimation (CE) • Elimination of Pilot overhead Limitations: • Low Spectrum Efficiency • Low SNR • Improved Bit Error Rate (BER)

3 Existing Methodology Block Type methodology which used all sub-carriers earmarked for pilots with a detailed period Comb Type methodology used certain sub-carriers is kept back for pilots for every symbol. Channel estimation algorithms for OFDM networks are investigated using a prototype setup. For medium estimate established on comb

162

S. Karimullah et al.

Fig. 1 Block Diagram of Existing Methodology

style pilot setup, different strategies for jointly estimating medium at pilots frequencies and integrating the medium are examined. At pilot frequencies, LS and LMS are employed to estimate the medium. Time Domain Interpolation is obtained by moving to a time domains by IDFT, minimal wadding, and then recovering to the frequency domain using DFT. Furthermore, medium estimation according to block kind pilot prearrangement is performed by transmitting pilots to every sub-channel and depending on this evaluation for a certain number of following codes. As illustrated in Fig. 1, we also built conclusion feedback equalisation for all sub-channels using intervallic block-type pilots.

4 Proposed Methodology With pilot spacing units 1, 2, and 3, the suggested blind CE is compared to comb as well as block type CE. You may alter the relationship between the vehicle speed as well as constellation size before executing the code. It displays the average throughput for every subcarrier as well as the BER in proportion to the SNR. When extensive DC biassing is used, the optical energy for each bit in relation to one side noise power spectral density, Ebopt/No, rises considerably, rendering the technique useless or unproductive. As a result, for one to preserve a low BER, DC bias has to be kept to a minimum; yet, since DC bias is significant, as mentioned above, an acceptable value of DC biassing is proposed. Effective communication between autos and traffic infrastructure (such as traffic signs, traffic lights, and so on) is crucial for guaranteeing the security of the traffic system. VLC is an exciting technology which might be the right solution to vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I) data transmission concerns. A significant issue in this respect is the construction of an appropriate sensor to decrease the influence of natural light during the day. Establishing good communication between cars and the transportation infrastructure,

Improved Spectral Efficiency Using Vehicular Visible Light …

163

which includes components like traffic signs, stoplights, and other roadside equipment, is essential to attaining this aim. Visible Light Communication (VLC) stands out as an exciting technology in this context that has the ability to overcome issues with data transfer between vehicles (V2V) and between vehicles and infrastructure (V2I). Information is transmitted with VLC using visible light, more especially the region of the electromagnetic spectrum that is visible to the human eye. Utilising current lighting infrastructure, such as LED lights, this technique modulates light signals’ intensities to communicate data. VLC allows for communication between cars and nearby infrastructure elements, resulting in the formation of a coherent and integrated network that improves overall safety. The generated signal is modulated by a light source and then transmitted through the WOC channel. After the DC bias is completely eliminated, the opposite operation is carried out at the receiver. Impulse reaction of the medium is squatter than Cyclic Prefix (CP). Deliberate fading properties of the channel is time-invariant compare to the symbol interval. Quadrilateral Windowing of the transmitted pulses Perfect Synchronization of transmitter and receiver additive noise, white noise, Gaussian channel noise. Furthermore, such findings would provide insight into how to select appropriate parameters for actual DCO-OFDM systems in order to improve the SE tradeoff. Finally, using a LOS link, it is shown that the stream of light approach perspective has a considerable influence on ophthalmic power circulation in the acceptance level as shown in Fig. 2. The VLC-OFDM signal, on the additional side, still has a maximum PAPR, and future research will focus on correcting this issue.

Initializatio n of the Parameters

Add Cyclic Prefix

Data Reshaping Process

FFT OFDM

Guard Band removal

Noise Effect

CE parameter Cyclic Prefix Computation removal

Fig. 2 Block diagram for proposed methodology

Output

Random Number generator

QAM Modulation

Symbol arrangement

CP for OFDM Symbols

164

S. Karimullah et al.

5 Results Software and Hardware Requirements Software used to simulate results is Matlab R2017b. Operating System requirements • • • •

W10OS W7 accommodated with SP1 PC accommodated with WS2019server PC accommodated with WS2016server

Processors With Minutest capacity of every Intel or AMD x86-64 processor However the endorsed processor is every Intel or AMD x86-64 processor contain of Four Logical Cores with AVX2 Instruction Set Facility Disk With Minutest capacity of 2.9 GB of HDD space for MATLAB only, 5–8 GB for a typical fitting However endorsed capacity of memory is an SSD is mentioned a full fitting of all Math Works products may take up to 29 GB of disk memory. RAM With Minutest capacity of 4 GB. The graph in Fig. 3 depicts the Spectral Efficiency vs. Ambient Light for a 16bit MQAM-OFDM communication system. This graph depicts how the system’s spectral efficiency varies depending on the ambient light conditions. The amount of information that can be transmitted per unit of bandwidth is referred to as spectral efficiency. According to the results in Fig. 3, the spectral efficiency is superior and optimal across the various ambient light conditions considered in Table 1. Table 1 provides a detailed comparison of spectral efficiency against various ambient light conditions. It includes scenarios such as a parking lot, lights off at night, lights on at night, partially cloudy but sunrise, sunny conditions, partially cloudy conditions, sunset, cloudy sunset, and shadowed environments. The table clearly shows that spectral efficiency remains consistently high and optimal despite the varying ambient light conditions. This demonstrates that the 16-bit MQAMOFDM system is robust and capable of maintaining performance regardless of the lighting conditions. Moving on, Fig. 4 depicts the Spectral Efficiency versus Zenith Angle for a 16-bit MQAM-OFDM system at a distance of 10 m in a parking structure. The zenith angle is the angle formed by the vertical direction and the line that connects the observer to the sun. The graph depicts how the spectral efficiency varies with zenith angle. The results show that the spectral efficiency remains superior and optimal across a wide

Improved Spectral Efficiency Using Vehicular Visible Light …

165

Fig. 3 Spectral efficiency versus ambient light (16-bit MQAM-OFDM)

Table 1 Spectral efficiency versus ambient light (16-bit MQAM-OFDM)

Ambient light

Spectral efficiency (bits/s/Hz)

a: Parking lot

1.16

b: Lights off in night

1.14

c: Lights on in Night

1.08

d: Partial cloudy but sunrise

1.02

e: When sunny

0.835

f: Cloudy

0.84

g: Partial cloudy

0.88

h: Sunset

1.06

i: Cloudy sunset

1.08

j: Shadow

1.00

range of zenith angles, indicating the ability of the 16-bit MQAM-OFDM system to maintain its performance under changing solar positions. Table 2 compares the spectral efficiency of the 16-bit MQAM-OFDM system in a parking structure at 10 m to different zenith angles. The table shows that spectral efficiency is consistently high and optimal across a wide range of zenith angles. This implies that the system can maintain its performance regardless of solar position or angle of incidence. In conclusion, both the spectral efficiency against ambient light conditions and the spectral efficiency against zenith angles show that the 16-bit MQAM-OFDM system performs superior and optimally. These findings demonstrate the system’s robustness and reliability under various lighting and solar conditions, highlighting its suitability for a variety of practical applications.

166

S. Karimullah et al.

Fig. 4 Spectral efficiency versus Zenith angle (16-bit MQAM OFDM) for parking structure at 10 m

Table 2 Spectral efficiency versus Zenith angle for 16-bit MQAM OFDM of parking structure at 10 m

ZE (deg)-Zenith angle

SE-Spectral efficiency (bits/sec/Hz)

−55

01.22

−45

01.34

−35

01.36

−25

01.4

−15

01.4

0

01.4

15

01.4

25

01.4

35

01.36

45

01.3

55

00.98

6 Conclusion The performance of VLC’s DCOOFDM schemes is evaluated in this research paper. The BER performance is reviewed as well as some issues of VLC design in practise. The simulation findings show that as similar with the DC biasing value grows, the BitErrorRate at the VLC scheme increases. When a strong DC biasing voltage is employed, the energy in optical medium per bit to Single Sided Noise Power Spectral Density, Ebopt/Number, increases dramatically, rendering the method inefficient or unproductive in terms of optical influence. In mandate to maintain a lower BitErrorRate, DC biasing mechanism essential be minimal; Although the existence of DC

Improved Spectral Efficiency Using Vehicular Visible Light …

167

bias can significantly affect a system, using a reasonable degree of DC bias is advised. The investigation of the impacts of SE-Ambient Light and SE-Zenith angle in a DCOOFDM (Direct Current Offset Orthogonal Frequency Division Multiplexing) system was the main goal of the research described in the previous sentence. The study’s results included an increase in the Signal-to-Noise Ratio (SNR) and a decrease in the Bit Error Rate (BER), which eventually resulted in an improvement in the system’s average throughput.

References 1. Siegel, J. E., Erb, D. C., & Sarma, S. E. (2018). A survey of the connected vehicle landscape— architectures, enabling technologies, applications, and development areas. IEEE Transactions on Intelligent Transportation Systems (TITS), 19(8). 2. Lakshmanna, K., Shaik, F., Gunjan, V.K., Singh, N., Kumar, G., Shafi, R. M. (2022). Perimeter degree technique for the reduction of routing congestion during placement in physical design of VLSI circuits. Complexity, 2022(Article ID 8658770), 11. https://doi.org/10.1155/2022/865 8770 3. Agarwal, S. (2022). IoT applications for health care. In Intelligent systems for social good: theory and practice (pp. 91–97). Springer Nature Singapore. 4. Rashid, E., Ansari, M. D., Gunjan, V. K., & Ahmed, M. (2020). Improvement in extended object tracking with the vision-based algorithm. In Modern approaches in machine learning and cognitive science: A walkthrough: Latest trends in AI (pp. 237–245). 5. Karagiannis, G., Altintas, O., Ekici, E., Heijenk, G., Jarupan, B., Lin, K., & Weil, T. (2011). Vehicular networking: A survey and tutorial on requirements, architectures, challenges, standards and solutions. IEEE Communications Surveys & Tutorials, 13(4), 584–616. 6. Gunjan, V. K., Kumar, S., Ansari, M. D., & Vijayalata, Y. (2022). Prediction of agriculture yields using machine learning algorithms. In Proceedings of the 2nd International Conference on Recent Trends in Machine Learning, IoT, Smart Cities and Applications: ICMISC 2021 (pp. 17– 26). Springer Singapore. 7. Sommer, C., & Dressler, F. (2014). Vehicular networking. Cambridge University Press. 8. Kwon, J., Ziegler, M. M., Carloni, L. P. (2019). A learning-based recommender system for autotuning design flows of industrial high-performance processors. In Proceedings of the 56th Annual Design Automation Conference 2019, Las Vegas, NV, USA, 2–6 June 2019, (pp. 1–6). 9. ETSI. (2013). “Intelligent Transport Systems (ITS); Vehicular Communications; GeoNetworking; Part 4: Geographical addressing and forwarding for point-to-point and point-tomultipoint communications; Sub-part 2: Media-dependent functionalities for ITS-G5,” ETSI, TS 102 636-4-2 V1.1.1, Oct. 2013. 10. Rudra Kumar, M., Pathak, R., & Gunjan, V. K. (2022). Machine learning-based project resource allocation fitment analysis system (ML-PRAFS). In: A. Kumar, J. M. Zurada, V. K. Gunjan, R. Balasubramanian (Eds.), Computational intelligence in machine learning. Lecture Notes in Electrical Engineering (Vol. 834). Springer. https://doi.org/10.1007/978-981-16-8484-5_1. 11. Chen, S., Hu, J., Shi, Y., Peng, Y., Fang, J., Zhao, R., & Zhao, L. (2017). Vehicleto-everything (V2X) services supported by LTE-based systems and 5G. IEEE Communications Standards Magazine, 1(2), 70–76. 12. Anandita Iyer, A., & Umadevi, K. S. (2023). Role of AI and its impact on the development of cyber security applications. In Artificial Intelligence and Cyber Security in Industry 4.0 (pp. 23– 46). Springer Nature Singapore. 13. Simsek, M., Aijaz, A., Dohler, M., Sachs, J., & Fettweis, G. P. (2016). 5G-enabled tactile internet. IEEE Journal on Selected Areas in Communications (JSAC), 34(3), 460–473.

168

S. Karimullah et al.

14. Nagaraju, C. H., Patil, M. K., Maheswari, C., Rahul, U. K., Rajesh, D. (2023). A novel wideband millimeter-wave-based OFDM uplink system to analyze spectral efficiency. In A. Kumar, S. Senatore, V. K. Gunjan (Eds.), ICDSMLA 2021. Lecture Notes in Electrical Engineering (Vol. 947). Springer. https://doi.org/10.1007/978-981-19-5936-3_74. 15. Manasa, K., & Leo Joseph, L. M. I. (2023). IoT security vulnerabilities and defensive measures in industry 4.0. In Artificial Intelligence and Cyber Security in Industry 4.0 (pp. 71–112). Springer Nature Singapore.

Modelling of Symmetrical 13 Level and Asymmetrical 31 Level Generalized Cascaded Multilevel Inverters Bolla Madhusudana Reddy, P. B. Chennaiah, and J. Chinnababu

1 Introduction MLIs are designed to provide industries with high-quality outputs and reduced ripples. Two-level inverters have suffered from several disadvantages, including high losses in the number of switches and greater power rating requirements. Typically, series combinations of switches are used to meet high voltage requirements, while parallel combinations are used for high current applications. However, this approach results in at least sixty percent Total Harmonic Distortion (THD), leading to increased power losses and decreased power quality output. High-current-carrying switches cause more power loss and temperature increase, leading to permanent damage. MLIs generate multiple voltage steps at the output with minimal harmonics, making them popular in recent times. MLIs find applications in high-voltage and high-power requirements. There are three main types of MLIs: (i) DC-MLIs, (ii) FC-MLIs, and (iii) CHB-MLIs. In addition to these classical MLIs, new topologies are available, such as modular MLIs and hybrid MLIs. Each type of MLI has its drawbacks. A 15-level non-insulated DC sources for novel symmetrical MLI with reduced switches has been developed to require fewer gate driver circuits and reduce complexity. A different topology B. M. Reddy (B) Associate Professor, Department of EEE, Malla Reddy College of Engineering for Women, Hyderabad, T.S., India e-mail: [email protected] P. B. Chennaiah Associate Professor, Department of EEE, Annamacharya Institute of Technology & Sciences, Rajampet, A.P., India J. Chinnababu Associate Professor, Department of ECE, Annamacharya Institute of Technology &Sciences, Rajampet, A.P., India © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 V. K. Gunjan et al. (eds.), Modern Approaches in Machine Learning and Cognitive Science: A Walkthrough, Studies in Computational Intelligence 1117, https://doi.org/10.1007/978-3-031-43009-1_16

169

170

B. M. Reddy et al.

for sub MLIs, serially connected, is proposed as a generalized MLI. The proposed MLI reduces switching devices in symmetric and asymmetric topologies. Various DC-link voltages are explained for asymmetric topologies, specifically the trinary providing a waveform with 3n levels, where n is the cells connected in cascaded, resulting in low THD and reduced dv/dt. A single-phase CHMLI is developed with a proposed basic unit in series with the H-bridge. Various algorithms are proposed to decrease the inverter’s cost by using fewer DC voltage sources, power switches, and installation space. An asymmetrical hybridized cascaded MLI is implemented for comparison with a symmetrical version, resulting in increased output voltage and double output load voltage for the asymmetric MLI DC sources. Simplified MLIs in symmetrical and asymmetrical forms are compared by calculating conduction losses and verified efficiency at various carrier frequencies. An asymmetrical type MLI has been validated based on Fuzzy, ANN and Indirect Field Oriented Control techniques. A hybrid MLI with an assisted boost network is proposed. Level shifted and phaseshifted PWM methods are evaluated for a 7-level inverter. The multicarrier PWM technique is implemented for a diodeless neutral clamped MLI. To achieve the desired amplitude voltage and frequency, the controlling technique is a crucial factor. Harmonics at low frequencies in VSIs produce power losses and pulsating torque in AC drives. Asymmetrical type MLIs with 31 levels and symmetrical 13 levels are implemented using a novel modulation method. Symmetric and asymmetric comparisons are made for low THD and with an asynchronous motor drive. The MC-LSSPWM control approach is used to provide an ideal organized 31-level MLI with the least number of switches and DC sources to improve the effectiveness of AC motor drives. By increasing the number of levels, the suggested MLI generates extremely low switching losses, minimal off-state voltage dips, and lowest THD owing to the sinusoidal discharge waveform.

2 Proposed Sub Multilevel Inverter Topology The suggested sub MLI has ‘n’ number of DC sources, as indicated in Fig. 1. Normally, the number of DC sources is uneven. Each DC source must be equal to Vdc to produce equivalent voltage steps. Each sub MLI has n+2 switches, some unidirectional and others bidirectional. Unidirectional switches are IGBTs with an opposing shunt diode. Switches S1, S’1, S(n+2)/2, and S’(n+2)/2 are unidirectional, while the remaining are bidirectional and endure positive and negative voltages. To achieve Vo = Vdc, switches S(n+2)/2 and Sn/2 are switched ON, and Vo=–Vdc is obtained with the ON condition of switches S(n−2)/2 and Sn/2. The same procedure is applied to the other switches. Bidirectional switches with two IGBTs and a single gate driver circuit generate positive and negative voltages. This sub multilevel structure provides positive or zero levels of output voltage. When switches S1 and S1 are both turned on, the output is zero.

Modelling of Symmetrical 13 Level and Asymmetrical 31 Level …

171

Fig. 1 Proposed sub MLI

The state of the switches for each output voltage is given in Table 1, where the value ‘1’ indicates the ON state, and ’0’ represents the OFF state of a switch. From Fig. 2, if two switches are required to be turned ON for every output voltage stage, the procedure is as follows: first, select a switch from the upper layer, and then select the second one from the lower portion. To obtain an output voltage of (n−1) * Vdc, the switches Sn/2 and S’(n+2)/2 should be switched ON, as shown in Table 1. Table 1 Output voltages and switching pattern of proposed sub MLI State

S1

S’1

S2

S’2

.

Sn/2

S’n/2

S(n+2)/2

S’(n+2)/2

Vo

1

1

1

0

0

.

0

0

0

0

0

2

0

1

1

0

.

0

0

0

0

Vdc

3

0

0

1

1

.

0

0

0

0

2Vdc

..

..

..

..

..

.

..

..

..

..

..

n−1

0

0

0

0

.

1

1

0

0

(n−2)Vdc

n

0

0

0

0

.

1

0

0

1

(n−1)Vdc

n+1

0

0

0

0

..

0

0

1

1

nVdc

Fig. 2 Single phase proposed generalized MLI

172

B. M. Reddy et al.

3 Proposed Generalized Cascaded Multilevel Inverter Topology The cascaded MLI suggested here is a series of sub multilevel inverters for attaining the needed voltage levels. ‘m’ is the total amount of sub MLIs in sequence in Fig. 2. Each sub MLI is made up of a ‘n’ number of DC voltage sources that are the same for every single sub MLI, whether symmetrical or asymmetrical. The sub MLI produces only positive and zero output voltage levels; to create negative levels, an H-bridge inverters is attached across the entire string of sub MLIs. Positive output voltages are acquired while the ON state of ‘T1’ and ‘T4’ devices, while negative output voltages are obtained during the ON position of ‘T2’ and ‘T3’ devices in the H-bridge.

4 Seven Level Multi Level Inverters Proposed symmetrical 7-level MLI has equal DC voltage sources for all levels as in the Fig. 3 with ‘R’ load of 100 Ω. To get +300 V the switches T1 , T4 , S3 , S2 , S1 required to ON. To get −300 V the switches T2 , T3 , S3 , S2 , S1 required to ON Table 2 shows ON state switching pattern for 7 level proposed symmetrical MLI. Fig. 3 Single phase seven level proposed symmetric MLI

Modelling of Symmetrical 13 Level and Asymmetrical 31 Level … Table 2 Switching pattern for proposed symmetrical 7-level MLI topology

Fig. 4 Single phase seven level existing cascaded H-bridge MLI

173

Voltage (Volts)

ON switching patterns

Level no

+300

T1 T4 S3 S2 S1

1

+200

T1 T4 S’3 S2 S1

2

+100

T1 T4 S’3 S’2 S1

3

0

S’1 S’2 S’3

4

−100

T2 T3 S’3 S’2 S1

5

−200

T2 T3 S’3 S2 S1

6

−300

T2 T3 S3 S2 S1

7

174 Table 3 Switching pattern for Hbridge symmetrical 7-level MLI

B. M. Reddy et al.

Voltage(Volts)

ON switching patterns

+300

S11 S34 S31 S24 S21 S14

Level no 1

+200

S11 S34 S33 S24 S21 S14

2

+100

S11 S34 S33 S24 S23 S14

3

0

S34 S33 S24 S23 S14 S13

4

−100

S12 S23 S24 S33 S34 S13

5

−200

S12 S23 S22 S33 S34 S13

6

−300

S12 S23 S22 S33 S32 S13

7

4.1 Conventional Seven Level Cascaded H-Bridge MLI The seven level existing H-bridge MLI is shown in Fig. 4 with ‘R’ load of 100 Ω. Toget +300 V, switches S11 , S34 , S31 , S24 , S21 , S14 required to ON. To get −300 V, switches S12 , S23 , S22 , S33 , S32 , S13 required to ON Table 3 shows ON state switching pattern for 7 level.

5 Modulation Method In this modulation technique, all triangular carriers are operated with equal frequency and magnitude with level shift. The number of carriers depends on the number of voltage levels, with the peak magnitude being the summation of all carrier wave magnitudes. The below Fig. 5 shows the modulation method for the 7-level proposed MLI, in which three high-frequency triangular carrier waves above zero are at levels of 100, 200, 300 V, and below zero at levels of −100, −200, −300 V, distributed along with a low-frequency sinusoidal reference wave. The firing pulses will be generated where the carrier wave magnitude is more than the reference wave.

Fig. 5 Multicarrier level shifted sinusoidal pulse width modulation for 7-level MLI

Modelling of Symmetrical 13 Level and Asymmetrical 31 Level …

175

Fig. 6 Proposed single phase symmetrical cascaded 13 level MLI

6 Symmetrical 13 Level Proposed Generalized Cascaded MLI The symmetrical cascaded MLI shown in Fig. 6 consists of all equal DC sources. In this MLI, two sub MLIs are implemented, each with three sources and five switches (IGBTs with anti-parallel diodes). An H-bridge with four switches is connected across the series of two sub MLIs. The type of load is RL with R = 100 Ω and L = 100 mH, with a total DC input of 300 V. Table 4 shows the ON state switching pattern for the single-phase 13-level MLI. To get an output voltage of +300 V (level 1), the ON state switches are S31 , T1 , T4 , S’22 , S32 , and S’21 . To get an output voltage of -300 V (level 13), the ON state switches are S31 , T2 , T3 , S’22 , S32 , and S’21 . The MC-LSSPWM is implemented with a sine wave as reference, and above zero, six triangular carrier waves and below zero, six carrier triangular waves are used for pulse generation to turn switches on.

7 Proposed Asymmetrical 31 Level Generalized Cascaded MLI The asymmetric cascaded MLI, DC voltage sources’ values differ from one sub MLI to other. In any one sub MLI, the DC sources are same but for another sub MLI another DC sources exist so that maximum levels in output is obtained, there by smooth wave form is achieved. In above n-DC sources number, m-sub MLIs number, Fig. 7 shows the proposed 31 level MLI constructed with four sub MLIs cascaded form across H-bridge connected.

176 Table 4 ON switching pattern of 13 level proposed symmetrical MLI

B. M. Reddy et al.

Voltage(V)

ON switching patterns

+300

S31 T1 T4 S’22 S32 S’21

Level no 1

+250

S31 T1 T4 S’22 S32 S’11

2

+200

S11 T1 T4 S’22 S32 S’21

3

+150

S31 T1 T4 S’12 S12 S’21

4

+100

S31 T1 T4 S’12 S12 S’11

5

+50

S11 T1 T4 S’12 S12 S’21

6

0

S’12 S12 S’11 S 11

7

−50

S11 T2 T3 S’12 S12 S’21

8

−100

S31 T2 T3 S’12 S12 S’11

9

−150

S31 T2 T3 S’12 S12 S’21

10

−200

S11 T2 T3 S’22 S32 S’21

11

−250

S31 T2 T3 S’22 S32 S’11

12

−300

S31 T2 T3 S’22 S32 S’21

13

Fig. 7 Proposed single phase asymmetrical cascaded 31 level MLI

Each sub MLI unit has one DC source with two switches. Table 5 shows ON state switches for respective voltage levels in output waveform. For example level 1st or +300 V requiredpurpose B’1 B’2 B’3 B’4 D1 D4 switches should be ON. Similarly B’1 B2 B3 B4 D2 D3 switches required ON for obtaining level 31st or −300 V voltage. The switches B1 B2B3B4should be conduct to get output of 0 V voltage or level 16th. In similar analysisapplicable for getting multiple voltage levels by turning ON

Modelling of Symmetrical 13 Level and Asymmetrical 31 Level …

177

Table 5 ON switching pattern of 31 level proposed asymmetrical MLI Voltage (volts)

ON switching patterns

Level no

Voltage (volts)

ON switching patterns

Level no

+300

B’1 B’2 B’3 B’4 D1 D4

1

−20

B’1 B2 B3 B4 D2 D3

17

+280

B1 B’2 B’3 B’4 D1 D4

2

−40

B1 B’2 B3 B4 D2 D3

18

+260

B’1 B2 B’3 B’4 D1 D4

3

−60

B’1 B’2 B3 B4 D2 D3

19

+240

B1 B2 B’3 B’4 D1 D4

4

−80

B1 B2 B’3 B4 D2 D3

20

+220

B’1 B’2 B3 B’4 D1 D4

5

−100

B’1 B2 B’3 B4 D2 D3

21

+200

B1 B’2 B3 B’4 D1 D4

6

−120

B1 B’2 B’3 B4 D1 D4

22

+180

B’1 B2 B3 B’4 D1 D4

7

−140

B’1 B’2 B’3 B4 D1 D4

23

+160

B1 B2 B3 B’4 D1 D4

8

−160

B1 B2 B3 B’4 D2 D3

24

+140

B’1 B’2 B’3 B4 D1 D4

9

−180

B’1 B2 B3 B’4 D2 D3

25

+120

B1 B’2 B’3 B4 D1 D4

10

−200

B1 B’2 B3 B’4 D2 D3

26

+100

B’1 B2 B’3 B4 D1 D4

11

−220

B’1 B’2 B3 B’4 D2 D3

27

+80

B1 B2 B’3 B4 D1 D4

12

−240

B1 B2 B’3 B’4 D2 D3

28

+60

B’1 B’2 B3 B4 D1 D4

13

−260

B’1 B2 B’3 B’4 D2 D3

29

+40

B1 B’2 B3 B4 D1 D4

14

−280

B1 B’2 B’3 B’4 D2 D3

30

+20

B’1 B2 B3 B4 D1 D4

15

−300

B’1 B’2 B’3 B’4 D2 D3

31

0

B1 B2 B3 B4

16

related switches. In this MLI, RL load used with values as R = 100 Ω and L = 100 mH, with total DC input of 300 V. In the proposed 31 level MLI, switches operated with MC-LSSPWM technique to obtain required levels at output. In this technique low frequency reference wave is sinusoidal waveform and high frequency carrier waves are triangular wave forms. In this 31 MLI, for positive half cycle of reference wave 15 number of carrier waves and for negative half cycle of reference wave 15 number of carrier waves are implemented at different positive negative magnitude levels respectively. Table 5 shows different voltage levels where carrier signals are distributed. Therefore above zero 15 positive levels, below zero 15 negative levels and zero level therefore 31 levels achieved.

8 Conventional Single Phase 31 Level Cascaded MLI The existing conventional single phase 31 level cascaded MLI topology shown in Fig. 8, in which four H-bridges are coupled in series through one DC source and 4 switches ineach H-bridge unit. Table 6 shows switching patterns with sources for obtaining voltage levels. Suppose if (20 + 40 + 80 + 160) volts voltage sources used, then +300 V achieved. Similarly for −300 V obtaining case, (–20–40-80–160) voltage sources need to operate. For example, if +20 voltage sources used means, the switches S12 , S13 should be turned ON and for −20 voltage source used case S11 ,

178

B. M. Reddy et al.

Fig. 8 Conventional single phase 31 level cascaded H-bridge MLI

S14 switches should be ON i.e., one from upper layer andanother from lower layer with opposite limbs. If anyone voltage source voltage not included case both the switches should be ON from upper layer and lower layer of same leg for gettingshort circuit. For example to obtain +100 V output, switches S32 S44 S42 S13 S12 S24 S22 S33 should be turned ON. For −100 V output case, required ON switches are S31 S43 S41 S14 S11 S23 S21 S34 . To obtain +300 V the ON state switches are S42 S13 S12 S23 S22 S33 S32 S43 . For −300 V output purpose the ON state switches are S41 S14 S11 S24 S21 S34 S31 S44 .In similar way, the remaining switches have been conducting sate for required voltage levels.

9 Modeling of Three Phase Induction Motor The conventional four-pole Induction Motor (IM) consists of two types of winding coils in every phase connected in series, and finally, the three phases are in a star formation, as shown in Fig. 9. The voltage profile of each half-winding coil in the proposed structure is equal to half of the voltage profile of the winding coil in the

Modelling of Symmetrical 13 Level and Asymmetrical 31 Level … Table 6 31 level conventional cascaded Hbridge MLI switching patterns with sources

179

Voltage (volts)

Switching patterns with sources

Level no.

+300

(20+40+80+160)

1

+280

(20+40+60)

2

+260

(20+80+160)

3

+240

(80+160)

4

+220

(20+40+160)

5

+200

(40+160)

6

+180

(20+160)

7

+160

160

8

+140

(20+40+80)

9

+120

(40+80)

10

+100

(20+80)

11

+80

80

12

+60

(20+40)

13

+40

40

14

+20

20

15

0

0

16

−20

−20

17

−40

−40

18

−60

(−20–40)

19

−80

−80

20

−100

(−20–80)

21

−120

(−40–80)

22

−140

(−20–40–80)

23

−160

−160

24

−180

(−20–160)

25

−200

(−40–160)

26

−220

(−20–40–160)

27

−240

(−80–160)

28

−260

(−20–80–160)

29

−280

(−40–80–160)

30

−300

(−20–40–80–160)

31

conventional arrangement. The voltage equations of the stator winding, which include resistance, self-inductance, and magnetizing inductance. The three individual single-phase 31-level MLIs are connected in a star formation; therefore, a three-phase optimal structure 31-level MLI is formed, which is connected

180

B. M. Reddy et al.

Fig. 9 Induction motor stat or winding: Proposed structure for MLI

with a three-phase induction motor to test the drive performance, as shown in Fig. 10. In the proposed three-phase 31-level MLI, each single-phase unit is connected to one phase with neutral grounded. Fig. 10 Three phase asymmetrical 31 level MLI fed IM drive

Modelling of Symmetrical 13 Level and Asymmetrical 31 Level …

181

10 Simulation Results and Analysis In the proposed Multi-Level Inverter (MLI) configuration, the approach involves the initial design of a sub MLI, followed by a series connection of all sub MLIs to create a generalized cascaded MLI. This structure is employed to compare and analyze the performance of various MLIs. The first comparison is made with a Seven-Level Symmetrical MLI, utilizing 10 switches. The corresponding voltage waveform is depicted in Fig. 11, demonstrating a THD of 34.60% (as shown in Fig. 12) and a fundamental voltage of 220.7 V. Additionally, the current waveform is presented (Fig. 13), revealing a THD of 30.15% (Fig. 14) and a fundamental current of 2.80 A. Contrasting this, an Existing Seven-Level Cascaded H-Bridge MLI is examined, requiring 12 switches. The voltage waveform is shown in Fig. 15, featuring a THD of 42.08% (as depicted in Fig. 16) and a fundamental voltage of 210.9 V. The associated current waveform is displayed (Fig. 17), indicating a THD of 32.16% (Fig. 18) and a fundamental current of 2.135 A. Further analysis includes a Cascaded 13-Level Symmetrical MLI, utilizing 14 switches and 6 DC sources. The output voltage waveform (Fig. 19) showcases a THD of 7.54% (Fig. 20) with a fundamental voltage of 298.2 V. Simultaneously, the

Fig. 11 Output voltage waveform of 7–level symmetric MLI

182

B. M. Reddy et al.

Fig. 12 THD in output voltage of 7-level proposed symmetric MLI

Fig. 13 Output current waveform of 7-level proposed symmetric MLI

Fig. 14 THD in output current of 7-level proposed symmetric MLI

output current waveform (Fig. 21) demonstrates a THD of 2.71% (Fig. 22) and a fundamental current of 2.869 amps. Notably, the Proposed Optimal Structured Cascaded 31-Level MLI requires 12 switches and 4 DC sources. The corresponding output voltage waveform (Fig. 23) displays a THD of 2.94% (Fig. 24), coupled with a fundamental voltage of 299.50 V. The output current waveform (Fig. 25) exhibits a THD of 1.12% (Fig. 26), along with a fundamental current of 2.914 amps.

Modelling of Symmetrical 13 Level and Asymmetrical 31 Level …

183

Fig. 15 Output voltage wave form of 7-level existing H-bridge cascaded MLI

Fig. 16 THD in output voltage of 7-level existing H-bridge cascaded MLI

Transitioning to the interaction with a three-phase Induction Motor (IM), the three-phase cascaded MLI is examined in the drive performance context. The threephase output voltage waveform of the 31-level MLI is visualized in Fig. 27, while the sinusoidal three-phase stator current waveforms (Fig. 28) are a result of increased waveform levels, leading to improved quality and harmonic suppression. The proposed optimal structured cascaded 31-level MLI-fed IM drive showcases consistent speed maintenance near the reference speed, superior torque in maximum and starting scenarios, and reduced rise and settling times (Fig. 28). Furthermore, when scrutinized under no-load and 5 Nm load conditions, the three-phase output

184

Fig. 17 Output current waveform of 7-level existing H-bridge cascaded MLI

Fig. 18 THD in output current of 7-level existing H-bridge cascaded MLI

Fig. 19 Single phase 13 level symmetrical cascaded MLI output voltage

B. M. Reddy et al.

Modelling of Symmetrical 13 Level and Asymmetrical 31 Level …

185

Fig. 20 THD in output voltage of single phase 13 level symmetrical cascaded MLI

Fig. 21 Single phase 13 level symmetrical cascaded MLI output current

Fig. 22 THD in output current of single phase 13 level symmetrical cascaded MLI

voltage demonstrates a THD of 2.94 and 8.33%, respectively, accompanied by fundamental voltages of 299.5 V. Correspondingly, the three-phase stator current exhibits a THD of 5.06% and a fundamental current of 2.851 A under the load condition. This enhanced performance of the induction motor through the utilization of the 31-level MLI is encapsulated in Table 9, which summarizes the three-phase 31-level MLI-fed IM drive’s performance under steady-state conditions (Figs. 29 and 30).

186

B. M. Reddy et al.

Fig. 23 Output voltage of single phase 31 level asymmetrical cascaded MLI

Fig. 24 THD in output voltage of the single phase 31 level a symmetrical cascaded MLI

Fig. 25 Output current of single-phase 31-level asymmetrical cascaded MLI

Modelling of Symmetrical 13 Level and Asymmetrical 31 Level …

187

Fig. 26 THD in output current of single-phase 31-level asymmetrical cascaded MLI

Fig. 27 Three-phase generalized cascaded 31-level MLI output voltage

As indicated in the preceding Table 7, it is evident that the proposed 7-level MLI exhibits several advantages, including a reduction in Total Harmonic Distortion (THD) and a decreased number of switches, all while achieving improved fundamental voltage and current characteristics at the same voltage level and number of voltage sources. In pursuit of even more enhanced attributes while minimizing the required DC sources, a symmetric 13-level and an asymmetric 31-level MLI were introduced. Notably, these configurations significantly lower the THD, as illustrated in the subsequent Table 8. In the above levels of MLIs the load is considered as RL. The single phase 31 level MLI has been intended as 3 phase 31 level MLI for verifying dynamic performance as shown in Table 9.

188

B. M. Reddy et al.

Fig. 28 Stator currents, speed, and electromagnetic torque for three-phase generalized cascaded 31-level MLI-fed Induction Motor a No-load case, b 5Nm Load case

Modelling of Symmetrical 13 Level and Asymmetrical 31 Level …

189

Fig. 29 THD of three phase voltage of 31 level cascaded MLI under NO load and 5 Nm load cases

Fig. 30 THD of three phase currents of 31 level cascaded MLI a NO load case, b ON load (5 Nm) case Table 7 Comparison between Proposed Single Phase 7-Level MLI and Existing 7-Level Cascaded H-Bridge MLI Type of MLI

Level No. of No. of % Fundamental % Fundamental no DC switches THD voltage (V) THD current (A) sources in in voltage current

Existing H 7 BridgeCascadedMLI

3

12

42.08

210.9

32.16

2.135

Proposed symmetrical MLI

3

10

34.60

220.7

30.15

2.80

7

190

B. M. Reddy et al.

Table 8 Comparison between proposed single phase symmetrical 13 level and asymmetrical 31 level MLI with DC input of 300 V & ‘RL’ load of R = 100 Ω, L = 100 mH Topology

DC input (V)

No. Of DC sources

No. of switches

%THD output voltage

Fundamental voltage (V)

%THD in output current

Fundamental current (A)

13 level symmetrical MLI

300

6

14

7.54

298.2

2.71

2.869

31 level asymmetrical MLI

300

4

12

2.94

299.5

1.12

2.914

Table 9 Three phase 31 level MLI fed IM drive performance under steady state condition Parameters

No load case

Load case (5 Nm)

DC input (Vph)

300

300

%THD output voltage

2.94

2.94

Fundamental voltage (V)

299.5

299.5

%THD in output current

8.33

5.06

Fundamental current (A)

1.854

2.851

Rise time tr (sec)

0.36

0.81

Settling time ts (sec)

0.40

0.90

Electromagnetic torqueTmax (Nm)

12

12

Speedm (rad/sec)

150

148

Load torque TL

0

5

11 Conclusion Initially, a single-phase symmetrical 7-level MLI is proposed with reduced switches and THD in comparison with the existing 7-level H-bridge MLI with the same DC sources. A single-phase 13-level symmetrical-type MLI is intended, and its output voltage, current waveforms, and THD are verified. Later, a single-phase asymmetrical 31-level MLI with an optimal structure has been built, and its voltage, current waveforms, and THD are verified. The proposed asymmetrical cascaded 31-level MLI has given minimum THD with reduced switches and DC sources than the 13level symmetrical MLI. We observed that the quality in voltage and current output increases with voltage level. The harmonics in the output current and voltage are suppressed with an increase in level. The proposed three-phase cascaded 31-level MLI is implemented with a star connection of three individual single-phase 31-level MLIs, and its output voltage, current waveforms, and THD are verified. Later, the three-phase 31-level MLI is fed with a three-phase IM, considering both without and with load conditions. In

Modelling of Symmetrical 13 Level and Asymmetrical 31 Level …

191

the proposed MLI-fed asynchronous motor drive, speed, torque, and stator currents are verified, demonstrating good performance. Hence, by observing the results, the harmonic content has been mostly minimized, and the performance of the drive is enhanced.

References 1. Rodriugz, J., Lai, J. S, & Pen, F. Z. (2002). Multilevel inverters, a survey of topology, control and applications. IEEE Transaction on Industrial Electronics, 49(4), 724–738. 2. Kumar, S., Ansari, M. D., Gunjan, V. K., & Solanki, V. K. (2020). On classification of BMD images using machine learning (ANN) algorithm. In ICDSMLA 2019: Proceedings of the 1st International Conference on Data Science, Machine Learning and Applications (pp. 1590– 1599). Springer Singapore. 3. Gopi, P., & Reddy, B. M. (2023). Optimal placement of DG and minimization of power loss using naked mole rat algorithm, ICT-PEP 23. International Conference on Technology and Policy in Energy & Electric Power, 23. 4. Kangarlu, M. F., & Babaei, E. (2013). A Generalized cascaded multilevel inverter using series connection of sub multilevel inverters. IEEE Transactions on Power Electronics, 28(2), 625– 636. 5. Reddy, B. M., Reddy, Y. V. S., & Kumar, M. V. (2019). Comparison of closed loop optimal high level novel multilevel inverter fed induction motor drive using PI controller and fuzzy logic controller. International Journal of Recent Technology and Engineering, 7(6). 6. Babaei, E., Laali, S., & Bayat, Z. (2015). A single-phase cascaded multilevel inverter based on a new basic unit with reduced number of power switches. IEEE Transactions on Industrial Electronics, 62(2), 922–929. 7. Dhanamjayulu, C., & Meikandasivam, S. (2017). Implementation and comparison of symmetricand asymmetric multilevel inverters for dynamic loads. IEEE Access, 6, 738–746. 8. Reddy, B. M., & Reddy, Y. V. S. (2022). Performance improvement of closed loop optimal cascaded high level multilevel inverter fed induction motor drive using ANFIS with low THD and effective speed-torque control. Journal of Electrical Systems, 18(1). 9. Reddy, B. M., Reddy, Y. V. S., & Kumar, M. V. (2019). Comparison of IFOC scheme of three phase optimal 63 level multilevel inverter connected induction motor using FLC and ANFIS. Journal of Mechanics of Continua and Mathematical Science, 14(3). 10. Bharatiraja, C., Sanjeevikumar, P., & Iqbal, A. (2019). Investigations of multi-carrier pulse width modulation schemes for diode free neutral point clamped multilevel inverter. Journal of Power Electronics (JPE), 19(3), 702–713.

Improved Radix-4 Fast Fourier Transform Algorithm Used for Wireless Communication J. Chinna Babu and K. Naveen Kumar Raju

1 Introduction The Discrete Fourier Transform (DFT) is well recognised for its importance in a variety of applications, which has inspired academics to delve into its fundamentals and create a slew of Fast Fourier Transform (FFT) algorithms [1]. The DFT is a fundamental mathematical technique that allows for the study and manipulation of frequency-domain signals. Its applications include signal processing, communications, image processing, audio processing, and others. The use of FFT and inverse FFT (IFFT) algorithms facilitates the computation of the DFT and its inverse [2]. Consider the following sequence of complex numbers: × 0… xN. The Cooley-Tukey approach stands out as the most extensively used FFT algorithm due to its consistent performance and simplicity [3]. These properties are extremely desirable when creating efficient FFT algorithm hardware and software implementations [4]. Notably, when compared to the Radix-2 FFT technique, the Radix-4 and Radix-8 FFT algorithms built from the Cooley-Tukey method show significant reductions in computational cost [5]. However, achieving additional reductions in computing complexity has been increasingly difficult, pushing academics to focus on twiddle factor evaluation improvements [6]. FFT algorithm improvements are critical in the creation of high-speed, low-power FFT processors used in wireless communications, radar systems, and other areas [7]. We built an effective Radix-4 method in our project by making modest changes to the traditional technique. These changes, which attempted to reduce the amount of twiddle factor assessments, resulted in reduced complexity and faster processing

J. C. Babu (B) · K. N. K. Raju Department of Electronics and Communication Engineering, Annamacharya Institute of Technology and Sciences, Rajampet 516126, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 V. K. Gunjan et al. (eds.), Modern Approaches in Machine Learning and Cognitive Science: A Walkthrough, Studies in Computational Intelligence 1117, https://doi.org/10.1007/978-3-031-43009-1_17

193

194

J. C. Babu and K. N. K. Raju

time. To accomplish this, we reindexed the output samples produced from the DFT decomposition [8]. Higher radix approaches, such as Radix-4 and Radix-8, can reduce difficult calculations, but also add complexity to the butterfly structure due to the presence of several input complex adders [9, 10]. Recent research has concentrated on the creation of multi-path pipelined FFT algorithms in order to improve their performance and efficiency. These endeavours seek to investigate novel approaches for improving performance characteristics in FFT implementations. Many architectures of Fast Fourier Transform processor are introduced to use in OFDM transmission, such as the Multipath Delay Feedback, single-path delay feedback, multipath delay commutator, Single Path Delay Commutator. Various architectures of Fast Fourier Transform (FFT) processors have been introduced in the context of OFDM (Orthogonal Frequency Division Multiplexing) transmission. These architectures try to conduct the FFT computations required in OFDM-based systems as efficiently as possible. The Multipath Delay Feedback architecture is one such example. This architecture includes various delay elements and feedback pathways that aid in concurrent and pipelined data processing. The usage of multiple routes allows for the parallel processing of several FFT blocks, increasing the FFT processor’s overall throughput. Another architecture is Single-Path Delay Feedback, which simplifies the design by employing only one delay element and feedback path. Although this architecture has a lesser throughput than the multipath variant, it is more efficient in executing FFT computations. Similarly, the Multipath Delay Commutator architecture is intended to process many FFT blocks at the same time. It uses a commutation mechanism to efficiently swap and transport data between several FFT processing units, enabling parallel processing and increasing the FFT processor’s throughput. The Single-Path Delay Commutator architecture, on the other hand, combines the benefits of a streamlined design (with a single delay element) with the commutation technique’s efficient data routing capabilities. This architecture achieves a good mix of complexity and performance, making it appropriate for several OFDM applications.These various FFT processor architectures offer trade-offs in terms of complexity, throughput, and resource utilisation. The architecture chosen is determined by specific application needs like as processing speed, available resources, and power limits.The MDF design is widely utilized as an answer to give a capacity rate of more than one GS/s among the many Fast Fourier Transform architectures. Many algorithms of FFT and dynamic scaling strategies are presented to lower the space and power consumption as shown in Table 1. This study addresses the need for effective and fast algorithms for performing the Discrete Fourier Transform (DFT). While the DFT is a fundamental mathematical procedure with many uses in signal processing, communications, image processing, and audio processing, existing algorithms may fall short of meeting the demands of real-time processing, resource-constrained systems, and demanding applications such as wireless communications and radar systems. Fast algorithms that can greatly reduce the computing time and complexity of the DFT while retaining accuracy and keeping the desired signal characteristics are in high demand. The goal of this project

Improved Radix-4 Fast Fourier Transform Algorithm Used for Wireless …

195

Table 1 History of Fast Fourier Transform Sl.No.

Year

Authors

Implementations

Limitations

1

1805

Carl Friedrich

Development of fast algorithms for DFT

Unpublished

2

1822

Joseph Fourier

Fourier and Harmonic Analysis, Fourier Transform

Did not analyse the computation time

3

1932

Frank Yates

Published the interaction algorithm

Cannot be applied to experiments with factors less than two levels

4

1965

James Cooley and John Tukey

Modern generic Fast Fourier Transform algorithm

Relies on Chinese remainder theorem

is to examine and develop novel fast DFT algorithms to overcome these issues and increase the efficiency of DFT computations in a variety of applications.

2 General Radix-4 Algorithm To increase the fastness of the operation Radix-4 algorithm is investigated. This can be achieved by changing the base to 4. If the base is increased for the same number, the power/index will drop. Since N = 43 (N = 4M), the number of steps for Radix-4 has been cut by 50%, resulting in only three stages. Radix-4 is a four-input, fouroutput system that uses an in-place algorithm. Every fourth-time sample, recursively the Radix-4 DIT-Fast Fourier Transform splits the DFT into four DFTs of the fourth length. The results of these shorter Fast Fourier Transforms are reused to compute a large number of outputs, lowering the total computational cost significantly. The Radix-4 algorithm is a method for increasing the efficiency and speed of Fast Fourier Transform (FFT) operations. It accomplishes this by modifying the algorithm’s base to 4, which allows for faster computing than the usual Radix-2 approach. The Radix4 algorithm has the advantage of reducing the amount of calculation steps necessary. When compared to the Radix-2 technique, the Radix-4 approach can cut the number of steps in half with a sequence length N = 43, which is a multiple of 4 (N = 4M). This step reduction results in only three stages being required for the Radix-4 FFT computation. The need for the Radix-4 algorithm in Fast Fourier Transform (FFT) operations arises from the demand for faster and more efficient computation techniques. The Radix-4 algorithm offers several advantages that address this need and contribute to its importance in various applications. 1. Increased Speed: The Radix-4 algorithm improves the speed of FFT computations compared to traditional methods like the Radix-2 algorithm. By changing the base to 4, the algorithm reduces the number of computation steps, resulting in

196

2.

3.

4.

5.

J. C. Babu and K. N. K. Raju

faster processing times. This speed enhancement is particularly beneficial in realtime applications where rapid data processing is essential, such as in wireless communication systems, audio and video processing, and radar systems. Efficient Resource Utilization: The Radix-4 algorithm utilizes an in-place computation approach, which means that it performs computations directly on the input data without requiring additional memory space. This efficient resource utilization reduces the need for additional storage and memory accesses, resulting in optimized hardware implementation and reduced overall computational cost. Computational Cost Reduction: The recursive nature of the Radix-4 algorithm allows for the reuse of intermediate results from shorter FFT computations. By reusing these results, the algorithm significantly reduces redundant calculations, leading to a substantial reduction in the total computational cost. This cost reduction is particularly important in applications that involve large data sets or require high-frequency resolution, as it enables efficient processing without sacrificing accuracy. Scalability and Flexibility: The Radix-4 algorithm is scalable and can be applied to FFT computations of varying sizes. It can handle sequences of lengths that are multiples of 4 (N = 4M) efficiently, making it adaptable to different system requirements. This scalability and flexibility are crucial in applications where the input data size may vary dynamically or where different FFT lengths are needed. Application Versatility: The improved speed and efficiency offered by the Radix4 algorithm make it suitable for a wide range of applications. It is commonly used in Orthogonal Frequency Division Multiplexing (OFDM) systems, which are widely used in wireless communication standards like Wi-Fi, LTE, and 5G. The algorithm’s fastness and low computational complexity enable efficient signal processing in these systems, leading to enhanced data transmission rates, improved spectral efficiency, and reliable communication.

In conclusion, the Radix-4 algorithm addresses the need for faster and more efficient FFT computations. Its speed enhancements, efficient resource utilization, computational cost reduction, scalability, and versatility make it important in various applications. By utilizing the advantages of the Radix-4 algorithm, systems can achieve accelerated signal processing, improved data transmission rates, and optimized resource utilization, ultimately leading to enhanced performance and efficiency in numerous fields. The Discrete Fourier Transform (DFT) is a fundamental mathematical operation that is used to analyse and transform frequency-domain signals. It has numerous applications in disciplines like as signal processing, telecommunication, picture processing, audio processing, and others. The DFT computes the spectrum representation of a discrete signal, exposing the frequencies and magnitudes present in the signal. As the demand for real-time signal processing grows, there is a need to develop more efficient and faster DFT methods. Traditional DFT techniques, such as the direct method, are computationally complex, making them unsuitable for real-time processing or resource-constrained systems. As a result, academics

Improved Radix-4 Fast Fourier Transform Algorithm Used for Wireless …

197

have concentrated on creating Fast Fourier Transform (FFT) algorithms that provide significant improvements in computational efficiency. The Radix-4 algorithm uses an in-place algorithm with a four-input, four-output system. This means that the input data is changed in the same memory region as the output data, which reduces the requirement for additional storage. The algorithm divides the DFT recursively into four smaller DFTs of one-fourth the length. These shorter FFTs are computed utilising the preceding stage’s data, resulting in a considerable reduction in total computational cost. The Radix-4 algorithm can efficiently compute a large number of outputs by reusing the results of the shorter FFT computations. This reuse of intermediate findings reduces superfluous calculations and increases the FFT process’s overall processing efficiency. To summarise, the Radix-4 technique speeds up FFT computations by increasing the base to 4. It uses an in-place approach to decrease the amount of computing steps and recursively separates the DFT into smaller FFTs. It successfully lowers the total computing cost by reusing intermediate findings. Because of these qualities, the Radix-4 technique is a good approach for accelerating FFT computations and boosting overall operation speed. Radix-4 Fast Fourier Transforms require just 75% as many complexes when compared to Radix-2 Fast Fourier Transform as shown in Fig. 1. By utilizing the results of minor, medium calculations to create large DFT frequency outputs, DIF-FFT and DIT-FFT for Radix-4 Fast Fourier Transform boost the speed of the calculation. N point input signals are decomposed in Radix-4 Fast Fourier Transform Algorithms. Every N/4 output is a sum of four input samples all multiplied by −1, j, or + 1,–j as shown in Fig. 2. General Radix-4 Butterfly is an indispensable building block for Fast Fourier Transform (FFT) computations, as it is generated by the Radix-4 algorithm. It is required for complex multiplication and addition in the frequency domain. Let’s examine the General Radix-4 Butterfly in greater detail. Fig. 1 General Radix-4 butterfly

RADIX-4 Butterfly

198 Fig. 2 Point Radix butterfly diagram

J. C. Babu and K. N. K. Raju

X (k)

X(0)

WN0

X (1)

X (k+1) WNk

X (2)

X (k+2) WN2k

X (3)

X (k+3) WN3k

General Radix-4 Butterfly accepts four complex inputs denoted by the letters X0, X1, X2, and X3. It calculates these inputs and produces four complex outputs labelled Y0, Y1, Y2, and Y3. The General Radix-4 Twiddle Factor Butterfly has twiddle factors, which are represented by Wk, where k is the index of the butterfly operation within the FFT. Twiddle factors are complex numbers used to incorporate phase shifts and amplitude scaling in butterfly computations. These factors are precomputed and saved to reduce computational costs during the FFT. The General Radix-4 Butterfly involves multiple computational procedures. a. Multiply and Add: X1 is multiplied by the initial twiddle factor, Wk0, and the resulting product is added to X0. – Add the result of multiplying X3 by the second twiddle factor, Wk1, to X2. b. Butterfly Operations: Y0 results from the addition of X0 and X2. – Determine Y1 by taking the difference between X0 and X2. – Adding X1 and X3 results in Y2. – Determine Y3 by taking the difference between X1 and X3. Additional Scaling: If necessary, apply scaling factors to the output values to normalise them.

Improved Radix-4 Fast Fourier Transform Algorithm Used for Wireless …

199

The Radix-4 FFT algorithm is intended to be reusable by the General Radix-4 Butterfly due to its recursive nature. It is applied iteratively to various phases of the FFT, enabling the computation of larger DFT sizes in an efficient manner. The outputs of one butterfly operation may be utilised as inputs for subsequent butterfly operations, thereby reducing the number of redundant computations. The General Radix-4 Butterfly is typically implemented in hardware implementations of the Radix-4 FFT using specific functional units such as multipliers, adders, and complex arithmetic units. These components allow the Radix-4 FFT method to perform multiple butterfly operations in parallel, thereby improving its computational efficacy. In FFT computations, the General Radix-4 Butterfly has a number of benefits: – Reduced computational complexity: The Radix-4 algorithm requires fewer computation steps than Radix-2 algorithms for the same DFT size, resulting in increased efficiency. – Increased parallelism: The butterfly operations can be parallelized, allowing FFT computations to be sped up by utilising hardware resources. Compatible with modern processors and digital signal processing (DSP) chips, the Radix-4 algorithm is particularly valuable for hardware implementations. In conclusion, the General Radix-4 Butterfly is an essential component of the Radix-4 FFT technique, allowing intricate multiplication and addition operations in the frequency domain to be performed quickly and efficiently. Recursion, reusability, and parallelism all contribute to the computational efficacy and performance of the Radix-4 FFT.

3 Improved Radix-4 Algorithm With the exception of a minor modification that has a substantial effect, Radix-4 DIF Fast Fourier Transform technique is usually similar to that of the disintegration strategy, and the twiddle factor matrix is the result of this. To refer to the Radix4 DIF Fast Fourier Transform algorithm, the modified Radix-4 DIF Fast Fourier Transform method will be utilised and that results from revised decomposition. The updated algorithm’s processing requires just four actual twiddle factors (cosine and sine) to be assessed or imported from the lookup database, it is shown that at the time of operating of modified algorithm only four true twiddle factors must be either assessed or added. The usage of six real twiddle factors is required by the basic use of standard Radix-4 DIF Fast Fourier Transform approach. As a result, the improved method reduces twiddle factor evaluations and lookup database requests by 33%. It’s worth mentioning that index generation saves money in the same way that the lookup table does. This improvement is obtained without raising the operational or by the design difficulties of the Radix algorithm. In comparison to the matching index (4n

200

J. C. Babu and K. N. K. Raju

Fig. 3 16-point Radix-4 Butterfly diagram

+ 3) of the standard Radix-4 DIF Fast Fourier Transform technique, the generated index (N + 4n-1) mod N shown in (5), since. The normal butterfly of an ordinary Radix-4 DIF Fast Fourier Transform technique, on the other hand, necessitates the use of six real twiddle factors. As a result, the updated approach saves 33% database queries. It’s worth noting that when the lookup table is used, the index generation saves money in the same way. This enhancement is made without increasing the algorithm’s computational or structural difficulties. In comparison to the usual Radix-4 DIF Fast Fourier Transform algorithm’s matching index (4n + 3), the production of new index (N + 4n − 1) mod N presented does not require any additional complexity. A typical 16 point Radix butterfly is shown as shown in Fig. 3 which contains 3 stages respectively.

4 Simulation Results These are simulation results obtained for our code for improvedRadix-4 algorithm. Simulation of the project code can be done in either Modelsim or Xilinx Vivado Softwares. Both the softwares have their own properties and are very efficient for the

Improved Radix-4 Fast Fourier Transform Algorithm Used for Wireless …

201

Fig. 4 Single waveform result

simulation. Xilinx’s Vivado software Package is a software used for both the synthesis and simulation of Hard ware Description Language designs, which replaces the Xilinx ISE software and adds functionality for a system on development of the chip and synthesis of the high-level designs. The entire design flow has been completely rewritten and rethought with Vivado. Modelsim is a Mentor Graphics multi-language environment for the hardware description language simulation including Verilog, Very High Speed Hardware Description Language and SystemC, as well as a built-in C debugger as shown in Figs. 4 and 5. ModelSim is compatible with Intel Quartus Prime, PSIM, Xilinx ISE, and Xilinx Vivado. Simulation can be done manually or automatically using a graphical user interface (GUI). Mentor’s HDL simulation tools, such as ModelSim PE and Questa Sim, come in a variety of editions.

202

J. C. Babu and K. N. K. Raju

Fig. 5 Multiple waveform results

5 Conclusion We suggested a new enhanced Radix-4 algorithm that reduced the amount of twiddle factor evaluations, resulting in a reduction in computational time and the best and fastest results. An improved Radix-4 method is proposed in this study. The look up table accesses and the decrement in the number of twiddle factor evaluations has been achieved by the typical decomposition of the Radix-4 FFT algorithm and it has been demonstrated by us in this paper. Without adding any extra calculational or design complexity, the above mentioned results were achieved. In this study we have presented the core notion for the enhanced Radix-4 algorithm can be extended to the algorithms of DIF and DIT FFT both with greater radices, and also for the Fast Fourier algorithms which are multidimensional. The number of the modules of ROM are reduced as a result of the proposed Radix-4 FFT algorithm in the hardware implementation of the architectures or the designs.

References 1. Jia, L., Gao, Y.,& Tenhumen, H. (1999). Efficient VLSI implementation of radix-8 FFT algorithm. In Proceeding IEEE Pacific Rim Conference, Communications, Computers and Signal Processing, (pp. 468–471). 2. Ma, Y. (1999). An effective memory addressing scheme forFFT processors. IEEE Transactions Signal Processing, 47, 907–911. 3. Ma, Y., & Wanhammar, L. (2000). A hardware efficient control of memory addressing for high-performance FFT processors. IEEE Transactions Signal Processing, 48, 917–921.

Improved Radix-4 Fast Fourier Transform Algorithm Used for Wireless …

203

4. Jiang, Y., Zhou, T., Tang, Y., & Wang, Y. (2002). Twiddle-factorbased FFT algorithm with reduced memory access. In Proceeding IEEE IPDPS’2002, April 2002 (pp. 70–77). 5. Babu, J. C., Rao, N. M., Ramana, K. et al. (2022). A dynamic hybrid decoder apprroach using EG-LDPC codes for signal processing applications. Wireless Personal Communications, 122, 1435–1454. https://doi.org/10.1007/s11277-021-08956-5 6. Srinivasan, S., Bai, P. S. M, Mathivanan, S. K., Muthukumaran, V., Babu, J. C., & Vilcekova, L. (2023) Grade classification of tumors from brain magnetic resonance images using a deep learning technique. Diagnostics, 13(6), 1153. https://doi.org/10.3390/diagnostics13061153 7. Gaddam, D. K. R., Ansari, M. D., Vuppala, S., Gunjan, V. K., & Sati, M. M. (2022). Human facial emotion detection using deep learning. In ICDSMLA 2020: Proceedings of the 2nd International Conference on Data Science, Machine Learning and Applications (pp. 1417– 1427). Springer. 8. Chowdary, M. K., Turaka, R., Alabduallah, B., Khan, M., Babu, J. C., & Kiran, A. (2023). Low-power very-large-scale integration implementation of fault-tolerant parallel real fast fourier transform architectures using error correction codes and algorithm-based fault-tolerant techniques. Processes, 11(8), 2389. https://doi.org/10.3390/pr11082389 9. Shaik, A. S., Karsh, R. K., Suresh, M., & Gunjan, V. K. (2022). LWT-DCT based image hashing for tampering localization via blind geometric correction. In ICDSMLA 2020: Proceedings of the 2nd International Conference on Data Science, Machine Learning and Applications (pp. 1651–1663). Springer. 10. Chang, Y. N., & Parhi, K. K. (2003). An efficient pipelined FFT architecture. IEEE Transactions Circuits Systems II, 50, 322–325 (June 2003).

Methodologies in Steganography and Cryptography–Review G. Krishna Murhty and T. Kanimozhi

1 1. Introduction The objective of steganography is to cover the actual presence of information exchange in unsuspecting digital media protections by embedding messages. The mechanisms of encryption, decryption and their usage in the protocols of communication are studied by cryptography or secret writing. In order to maintain security of information, both approaches transform data, although the principle of steganography is different than cryptography. Cryptography distorts an important message, but it does not hide the fact that there exists a message. The objective of cryptography is to render third party data unreadable, while steganography is intended to hide third party data. Both approaches come from an older year, although the contemporary field is rather young. The essential components of computer security are cryptography and Steganography. Cryptography is an established mathematical basis of computer safety and an area of computer science that is well developed and actively explored [1]. 1. Applications of Steganography • To secretly transmit data over an uncertain channel. • To provide security against data change • Can be used in TV, audio & video broadcasting parallelly. 2. Various technologies in Steganography G. K. Murhty (B) · T. Kanimozhi Department of Electonics and Communication Engineering, Veltech Rangarajan Dr. Sagunthala R & D Institute of Science and Technology, Avadi, Chennai, Tamil Nadu, India e-mail: [email protected] T. Kanimozhi e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 V. K. Gunjan et al. (eds.), Modern Approaches in Machine Learning and Cognitive Science: A Walkthrough, Studies in Computational Intelligence 1117, https://doi.org/10.1007/978-3-031-43009-1_18

205

206

G. K. Murhty and T. Kanimozhi

I. Text Steganography The text steganography may be categorized as embedding in character level, bit level and mixed type into three main classes based on the typology of embedding technology. In Fig. 1 all these categories and their associated subcategories are indicated [2]. II. Audio Steganography Audio Steganography includes the audio as carrier message and secret message can be any of text, audio or image. The transmission of audio data is a difficult problem, the little modification of voice/speech/audio data reflects in a very significant way. It appears to be an excellent carrier as it contains considerably more redundant data [3] as in Fig 2. III. Video Steganography This method can hide secret information behind a video clip in order to cover a huge volume of data, as demonstrated in Fig. 3. IV. Image Steganography Fig. 1 Subcategories in text steganography

Fig. 2 Sample block of audio steganography

Methodologies in Steganography and Cryptography–Review

207

Fig. 3 Sample block of video steganography

Fig. 4 a Original image, b Secret image + Original image

The secret data underneath any cover image may be buried with this approach. In textual or visual form hidden data exist. This may be sent via an insecure channel after the stego image has been inserted as shown in Fig 4.

2 2. Steganography versus cryptography 1. Steganalysis Steganalysis is the technique of analyzing steganography technologies in which the concealed data in a stego item is detected, extracted, destroyed and manipulated. Attack may be of many sorts, for instance, some attacks just confirm the presence of hidden data, others try to discover and retrieve the hidden data, and some aim to destruct the hidden data by identifying the existence without trying to retrieve hidden data shown in Table 1.

208

G. K. Murhty and T. Kanimozhi

Table 1 Comparison between Steganography & Cryptography S. No

Steganography

Cryptography

1

Steganography implies writing covered

Cryptography implies writing secretly

2

Steganalysis is the name of the attack in Steganography

The name of Attack is Cryptanalysis, while in cryptography

3

Data structure is usually not changed in steganography

The structure of the data is changed during cryptography

4

The fact that a secret message occurs is covered in steganography

Only secret message is hidden in cryptography

5

Implemented on audio, video, image and text

Implemented only on text files

The detection of steganography itself is sufficient for uncovering even if the secret message is not retrieved since it is sufficient to identify the presence of concealed data for destroying it. The detection of certain characteristics of images that are affected by the concealed data is often carried out via identification. A skilled Steg analyst has to know the steganography instruments’ methodologies and tactics for effective attacks [4]. Some of the attacks will be: i. Stego only attack: For analysis, only stego objects are provided. ii. Known cover attack: Cover and stego are both familiar. iii. Known message attack: Message is known in some situations and the analysis of the stego object pattern can assist attack comparable systems for this embedded message. iv. Chosen stegoattack: Steganographic and stego object algorithms are known v. Chosen message attack: Here, Steg analyst produces example stego items from different steganographic tools to analyze the selected message and the algorithm employed is analyzed. vi. Known stegoattack: The object covers and technique used in steganography are known. 2. Cryptanalysis Any form of attacker should be resisted by a secure encryption system; otherwise, secure communication should be unable to utilize the encryption system. The focus of cryptographic analysis is on the plaintext recovery despite obtaining the secret key. The following are detailed four common forms of cryptosystem security analysis attacks [5]: i. Ciphertext-only attack: In order to get knowledge on the plaintext or the secret keys, the attackers do not know the encryption method and may only test it on the basis of a number of ciphertext intercepts. ii. Known-plaintext attack: Besides the intercepted ciphertext, there are known plaintext-ciphertext pairings for cryptosystem decryption.

Methodologies in Steganography and Cryptography–Review

209

iii. Chosen-plaintext attack: The attackers are able not only to select or construct certain plaintexts and their matching cyphertexts to improve the attacking performance iv. Chosen-ciphertext attack: The attackers may select or generate certain ciphertexts and acquire the appropriate complaints since they have the possibility to access the decryption process.

3 Methodologies In the article “An Efficient Tagged VisualCryptography for ColorImages” author implemented a tag stamping method for color images. The tag secret adds to the protection to the tagged shares.Tagging is a simple operation that allows good quality pictures to be produced. The suggested secret-sharing method, marked in color, distributes the secrets of the tags and replants it with the tag secret.The secrets created by stacking the shares also lead to a strong visibility of the secret. This may be expanded further using standard matrix-based VC or probabilistic VC for numerous secret pictures [6]. In the article “Image Encryption using Secure Force Algorithm with Affine Transform for WSN” author provided a symmetric technique of low complexity labeled as Secure Force with Affine Transform. The encryption component may be carried out using a minimal architecture consisting simply of fundamental mathematical operations (AND, OR, XOR, XNOR, SHIFTING, SWAPPING). This can also assist to decrease the strain on the encoder because only the decoder performs the most difficult key expansion procedure. The aim of this study is to analyze the safety and performance of the proposed approach. Using histogram, MSE and PSNR metrics, the performance of the technology may be assessed. The author analyzed and suggested that this technique provides higher PSNR and less MSE values [7]. In the article “New Watermark Embedding Technique using Visual Cryptography” author presents the brief notion of encryption and watermarking on images. This is why they are creating n number of shares in the secret image of progressive visual cryptography (PVC). PVC implies the gradual recovery of the hidden picture by overlaying n parts of a given secret image.They have the following scheme structure to share a secret image. This structure of the scheme maybe offers matrices that show how secret images can be shared across n shares. The structure of the scheme is designed in Matlab. The watermark embedding approach is applied here after the prediction error utilizing these pixels is found with the assistance of determining different sorts of pixels. This work can’t extract the watermark (secret data) from watermarked image (embedded image) at the same time the cover and the hidden image cannot be revealed [8]. In the article “Preserving Privacy Using Visual Cryptography in Surveillance Systems” a solution to the visual cryptographical monitoring system is implemented to save individual personal data from original imagewhich is a cost-effective and confidential. The algorithm of facial detection detects the faces of the original true

210

G. K. Murhty and T. Kanimozhi

image and then random data replaces the detected faces to protect the identity of the individuals. The other image is maintained, as it can track the actions within the monitoring system. Whenever a suspected act takes place, independent organizations in the system can work together to reproduce the first image and to identify the culprit. The suggested framework is based on XOR operations which are popular or even available as logic gates in the most basic processor architecture. The performance problems with XOR operations are thus disregarded. Face Detection techniques may cause the most serious difficulties in performance phase [9]. In the article [10] author examined the existing of Classical Cryptography (CC) and Quantum Cryptography (QC) works and principles utilized for encryption and decoding. The strengths and limitations of each encryption and decryption technology is discussed. The author focuses in this work on selecting the optimal approach for encrypting and decrypting images, so that researchers might have an idea of selecting an effective cryptography. Classical Cryptography transforms plain text into encrypted, i.e., cipher, to enable it to be sent across insecure exchanges of information. A data string called “Key” controls the information translation from not secured content to encrypted content. Quantum cryptography is the safest method of data transmission since it is centered on quantum physics principles, which allow both parties to secure communications centered on quantum mechanics’ invulnerability. The mathematical framework or set of principles used to build physical theory is the quantum mechanics. The Heisenberg Uncertainty Principle and the Photon Polarization Principle are two fundamental features of quantum cryptography that are reliant on quantum cryptography. Finally, quantum cryptography is certainly safer than that in traditional encryption, as it works on quantum mechanics principles and the characteristics of quantum bit approaches. Quantum encryption is particularly beneficial for safe sharing of multimedia data such as clips and photos. The Quantum Cryptography is in the beginning phases and much effort has to be completed in order to improve its performance and to overcome its limits such as the issues of its implementation, communication range and transfer rate [10]. In the article “Visual Cryptography for Color Images using Multi-level Thresholding”author in this work uses multi-level thresholding/halftoning to enhance the acquired quality of the reproduced image to implement a visual cryptography approach for colors images. Previous approaches using binary halftoning can only have 2 variations of any color. Shade 1 is the maximum intensity color of 255 while Shade 2 is its modestpower of 0.The fmultilevelhalftoning approach is superior than binary halftoning as it allows the reconstructed image to contain more color shades or more information levels than the binary halftoningmethodology.Multilevel halftoning at 5 level reaches 15 levels of information, which would rise only further when the halftoning levels are increased in binary halftoning compared to the always-fixed five levels [11]. In the article “An Extended Visual Cryptography Technique for Medical Image Security” studies of the author offer an expanded technique for the security of medical images in visual cryptography. Visual encryption is a way to share information secretly in forms such as images, text, etc. The approach suggested in the study encrypts the medical image first and then incorporates it into 3 cover images. The

Methodologies in Steganography and Cryptography–Review

211

secret image will be rebuilt later on the receiver side, followed by its decryption, using three shares (meaningful). For every pixel of the secret image, the significant shares utilized in the approach described employ a 3 × 3 block size. The study proposes no pixel expansion method for encryption. This method’s performance is evaluated with the SSIM, MSE and PSNR metrics. The transparency of secret images in the shares is high and attackers can change the secret information in shares [12]. In the article “Steganography in Images Using LSB Techniques” author presents a new technology for data hiding based on LSB digital image technology. This study presents an unsustainable data hiding approach utilizing LSB in images. In every media: text, sound, video and images, secrets may be encoded. His article addresses the use of Least Significant Bit (LSB) technology for hiding texts in an image file. The LSB method is applied in the spatial domain, where the carrier bits are included in the least significant bits of the image to generate the image of the stego. The major subject of this work is to embed a message in the image via a secret key or password into the stego system encoder. It should be kept private using this password or secret key. This stego image is sent to the recipient through a channel. With the same key or password, the stego system decodes the stego picture at the ending of the decoder. This work will have low embedding capacity and less in secrecy maintenance [13]. In this article “Holo-Entropy and AdvancedEncryption Standard for WaveletBased ImageSteganography”, a technique is used in this work to hide the sound information in the image that uses a Discrete Wavelet Transform (DWT).The secret message, which is included in audio, will initially be normalized and encrypted using the Advanced Encryption Standard (AES) technique. The encrypted message is then transformed into binary data that is inserted into the cover image. Using the Wavelet Transform, the coefficients are converted and chosen for optimum embedding using the holo entropic function. In the encrypted audio data, the specified coefficients are utilized. Lastly, the embedded band has to be converted and compressed with JPEG to the spatial domain. The similar &opposite technique is used on the receiver side for the extraction of the audio message. The suggested approach of image steganography with the expense of holo entropy and the AES algorithm is assessed based on MSE and PSNR [14]. A text steganography approach in JPEG pictures is provided in this article “A New Steganography Method for Embedding Message in JPEG Images”. Text steganography is done on the least bits in the discrete matrix, and the insertion of embedded message has a lesser influence on the image quality. The encoded message in the image is hidden with two less significant bits of pixels. In the JPEG compression method before encoding, after the transformation from time to frequency space and exactly (or immediately) following the discretion of converted data the steganographic operation is done to the image. The suggested approach is evaluated on the collection of several well-known images using two standard steganography criteria, PSNR and capacity [15]. In the article “Secured Communication of Text and Audio using Image Steganography” author suggested a securing text- and audio-steganographic communication utilizing an image-steganographic technique provides text and audio data security in a single image file. Audio recordings, as well as text files here, are transformed

212

G. K. Murhty and T. Kanimozhi

into binary and encoded using bitwise encoding into the cover file. For dual security method the LSB algorithm is utilized. In the first step the hidden text in the cover image is covered, but in the second phase the cover image file is saved with the secret audio file. The system presented was more efficient using the approach of transforming wavelets. The key is separately supplied here for encoding and decoding but it can’t store the key in the image randomly. The final or stego image that results is delivered to the receiver over the specified communication channel [16]. In this article “LSB Based Steganography using Dynamic Key Cryptography” author proposed for steganography of the space domain image. In this situation, the secret image’s most significant bit (MSB) is substituted for the image element carrier image’s least significant bit (LSB).The dynamic key cryptography provides security here, together with information hiding. This approach is designed to generate three sets of pseudo noise sequences that conceal the most significant bits (MSB) in each pixel of the private image from least significant bits (LSB) in each pixel of the carrier image. On utilizing the 192-bit key-vector we can construct a pseudo-noise sequence and it provides adequate security to implement the technique. Each block here is ciphered with an 8-bit character second key. If any attacker does a stego image analysis, it is only possible to get encrypted text. The four least significant bits (LSB) of each pixel of the carrier images are replaced with the four most significant bits (MSB) of all secret image pixel elements in the LSB steganography technique. Performance is evaluated based on PSNR [17]. This study, “Image and Text Steganography with Cryptography using MATLAB” combines cryptographical and steganographic security and has created a heavily secured model. Sequential algorithms are employed in this work for steganography and for cryptography, symmetric XOR is used. This article shows how to hide a text document in an image file and how to hide an image in another image. The app will initially be developed on the cloud and the input is provided via android, when the images from the android smartphone go to the cloud server where Matlab processes the images and results are delivered to the android smartphone. This work is only applicable to text and images files, not for video and audio files [18]. In this study “The author of “MP4 Steganographic in the Wavelet Transfomr Domain” suggests a Mp4 steganography solution that uses the wavelet domain to disguise audio and visual data. A Video file contains an I frame, a P frame, and a B frame. An iframe or intra frame is decoded without the need for extra frames to be referenced. Capturing the motion is done using P frames and B-frames. The GOP is made up of a number of frames, including I-, P-, and B- frames. A GOP begins with only an I-frame, which varies in length depending on the codec used (excluding the second I-frame, the number of frames between two I-frames). They are chosen for integration because I frames through compression or other types of signal processing techniques are not lost. For embedding, appropriate I-frames and audio frame pixels are used without affecting the video’s perceptual qualities. To avoid the loss of frames, redundant copies of secret data are produced. The audio frame mask the secret sound/audio, whereas the frames conceal the hidden visuals. The integer wavelet transformation is used to convert the concealed image and audio data, and the low frequency coefficients remain hidden [19].

Methodologies in Steganography and Cryptography–Review

213

In this article “Secure Video Steganography Technique using DWT and H.264”, a safe calculation of the video steganography using Discrete wavelet transform (DWT), based on the calculation of the movement object, and H.264, the use of Mat laboratory programming is presented. By utilizing bit shifting and H.264 to encode secret details, the secret message is prepared beforehand. In order to recognize local excitement for the moving articles, moving object detection calculation will first be updated with the host records. At that stage, it is the job of the information dissimilar process to insert the secret message image to the discrete wavelet that transforms the planes of all motion zones on the base of the video [20].

4 Conclusion This study mainly focused on introducing several approaches of steganography and cryptography. As we stated previously, both cryptography and steganography offer different data protection characteristics across the network. But they have not given accurate results if they are not utilized in combination form. LSB is the most used technology, however, but it has so many disadvantages that it decreases image quality and also generates suspicions. And a very few works where found on video processing of both steganography and cryptography. So, in future we can implement a technique based on steganography and cryptography in video processing which is more effective than pre-existing works.

References 1. Kortsarts, Y., & Kempner, Y. (2015) Steganography and cryptography inspired enhancement of introductory programming courses. InformationSystems Education Journal (ISEDJ), ISSN: 1545-679X. 2. Krishnan, R. B., Thandra, P. K., & Baba, M. S. (2017) An overview of text steganography. In 2017 Fourth International Conference on Signal Processing, Communication and Networking (ICSCN) (pp. 1–6). https://doi.org/10.1109/ICSCN.2017.8085643. 3. Patil, A. S., & Sundari, G. (2018). An embedding of secret message in audio signal. In 2018 3rd International Conference for Convergence in Technology (I2CT) (pp. 1–3). https://doi.org/ 10.1109/I2CT.2018.8529549. 4. Ansari, M. D., Gunjan, V. K., & Rashid, E. (2021). On security and data integrity framework for cloud computing using tamper-proofing. In ICCCE 2020: Proceedings of the 3rd International Conference on Communications and Cyber Physical Engineering (pp. 1419–1427). Springer. 5. Karimullah, S., Vishnuvardhan, D., & Bhaskar, V. (2022). An improved harmony search approach for block placement for VLSI design automation. Wireless Personnel Communications, 127, 3041–3059. https://doi.org/10.1007/s11277-022-09909-2 6. Talukdar, J., Singh, T. P., & Barman, B. (2023). Learning evaluation for intelligence. In: Artificial intelligence in healthcare industry. Advanced Technologies and Societal Change. Springer. https://doi.org/10.1007/978-981-99-3157-6_11. 7. Lakshmanna, K., Shaik, F., Gunjan, V. K., Singh, N., Kumar, G., & Shafi, R. M. (2022). Perimeter degree technique for the reduction of routing congestion during placement in physical design of VLSI circuits. Complexity, 2022, 1–11.

214

G. K. Murhty and T. Kanimozhi

8. Krishna, S. L. V., Abdul Rahim, B., Shaik, F., & Soundara Rajan, K. (2010). Lossless embedding using pixel differences and histogram shifting technique. In Recent Advances in Space Technology Services and Climate Change 2010 (RSTS & CC-2010), Chennai, India, (pp. 213–216). https://doi.org/10.1109/RSTSCC.2010.5712850. 9. Karimullah, S., Vishnuvardhan, D., Arif, M., & Gunjan, V. K. (2022). Research article an improved harmony search approach for block placement for VLSI design automation. 10. Talukdar, J., Singh, T. P., & Barman, B. (2023). Tools and technologies for implementing AI approaches in healthcare. In Artificial intelligence in healthcare industry. Advanced Technologies and Societal Change. Springer. https://doi.org/10.1007/978-981-99-3157-6_10. 11. Palsodkar, P., Palsodkar, P., Gokhale, A., Dorge, P., & Gurjar, A. (2022). Fuel larceny and leakage indication system using IoT. In Intelligent systems for social good: theory and practice (pp. 81–89). Springer Nature Singapore. 12. Rudra Kumar, M., Pathak, R., & Gunjan, V. K. (2022). Machine learning-based project resource allocation fitment analysis system (ML-PRAFS). In Computational Intelligence in Machine Learning: Select Proceedings of ICCIML 2021 (pp. 1–14). Springer Nature Singapore. 13. Singh, A. K., & Singh, J. (2015). Steganography in images using LSB technique. International Journal of Latest Trends in Engineering and Technology (IJLTET), 5(1). ISSN: 2278-621X. 14. Waghmare, W., Nipanikar, S. I., & Mulmule, P. V. (2018). Holo-entropy and advanced encryption standard for wavelet-based image steganography. International Journal of Research in Engineering, Science and Management, 1(10). www.ijresm.com|ISSN (Online):2581-5782. 15. Mukherjee, S., Nath, S. S., Singh, G. K., & Banerjee, S. (2022). FACEIFY: Intelligent system for text to image translation. In Intelligent systems for social good: Theory and practice (pp. 51–62). Springer Nature Singapore. 16. Anusha, M., Bhanu, K. N., & Divyashree, D. (2020). Secured communication of text and audio using image steganography. In 2020 International Conference on Electronics and Sustainable Communication Systems (ICESC) (pp. 284–288). https://doi.org/10.1109/ICESC48915.2020. 9155715. 17. Patel, N., & Meena, S. (2016). LSB based image steganography using dynamic key cryptography. In 2016 International Conference on Emerging Trends in Communication Technologies (ETCT), (pp. 1–5). https://doi.org/10.1109/ETCT.2016.7882955. 18. Saritha, M., Khadabadi, V. M., & Sushravya, M. (2016). Image and text steganography with cryptography using MATLAB. In 2016 International Conference on Signal Processing, Communication, Power and Embedded System (SCOPES) (pp. 584–587). https://doi.org/10. 1109/SCOPES.2016.7955506. 19. Hemalatha, S., Acharya, U. D., & Shamathmika. (2017). Mp4 video steganography in wavelet domain. In 2017 International Conference on Advances in Computing, Communications and Informatics (ICACCI) (pp. 1229–1235). https://doi.org/10.1109/ICACCI.2017.8126010. 20. Renuka, B., & Manja Naik, N. (2019). Secure video steganography technique using DWT and H.264. In 2019 1st International Conference on Advances in Information Technology (ICAIT) (pp. 19–23). https://doi.org/10.1109/ICAIT47043.2019.898740.

Study of Secure Data Transmission-Based Wavelets Using Steganography and Cryptography Techniques K. Ravindra Reddy and Vijayalakshmi P.

1 Introduction To enhance the data transmission security in network, steganography along with cryptography are utilized mutually by means of wavelets described by several researchers. Basically, in discrete wavelet transform “Filter bank investigation” might be utilized for examining signal images by passing all the way through it. The covert information which is surrounded in carrier that may help to protected from malevolent users sending to receiver based on steganography. Both steganography and cryptography approaches were introduced to covert the top secret message in the form of audio, video or image files. The principles utilized in cryptography during data transmission are mentioned as follows. (1) Encryption and Decryption technique-Converting original information into incomprehensible format hence the data is very secure which comes under encryption method. And then the cipher text is coverted into original form using decryption process to retrieve the original data. Either similar keys or dissimilar keys are utilized for both encrypting and decrypting the data. (2) Validation/Authentication-Validation is one of the major part in cryptography algorithm that the message was started by the originator stated in the message. This is possible if the sender performs some operation on the message that the recipient only knows the message originator may perform. (3) Reliability. K. Ravindra Reddy (B) · P. Vijayalakshmi (B) Department of ECE, Vels Institute of Science, Technology and Advanced Studies, Pallavaram, Chennai, India e-mail: [email protected] P. Vijayalakshmi e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 V. K. Gunjan et al. (eds.), Modern Approaches in Machine Learning and Cognitive Science: A Walkthrough, Studies in Computational Intelligence 1117, https://doi.org/10.1007/978-3-031-43009-1_19

215

216

K. Ravindra Reddy and P. Vijayalakshmi

The loss of message reliability is a prevalent concern in communication, which implies that using cryptographic hashes, cryptography should ensure that the messages received by the recipient are not altered anywhere along the communication path. (4) Without negation. This is a condition in which the sender may claim that he or she did not transmit the message to the intended recipient. With the use of computerized marks, cryptography maintains this type of situation. The benefit of utilizing the combination of both cryptography and steganography is to transfer the data very secure among transmitter and receiver side. The main objective of steganography is to remain the availability of covert data when broadcasting message in network. Also the objective of cryptography is to send the data as hidden information to transmit the data in a secure manner. Hence we surveyed how steganography and cryptography techniques used in encrypt/decrypt the message. To transfer the data with protection during data transmission. This survey focused on secure data transmission using steganography and cryptography are described as follows. • To send the data/information from sender to receiver using cryptography and steganography in a secure manner • To improve the security of information during message sending as hidden data which is embedded with either image file or audio or video file. The contribution of this survey is to construct safe and also undetectable passing of data, we introduced this survey for highlighting how secure transmission had done via cryptography and steganography based technique. The plain text is embedded with secret key along with secret message as stego image to secure the data from hacker who is stealing data.

2 Background Vasanth et al. [1] and Laskar et al. [2] introduced steganography based Joint Photographic Expert Group along with exchange encryption method. Moreover this JPEG method utilized Discrete Cosine Transform approach that utilized in occurrence area for concealed encrypted message inside the images. This proposed work was evaluated by measuring metrics such as Peak Signal Noise Ratio and Mean Square Error. Marwa et al. [3] A combined approach for data protection has been introduced in this paper which utilizes cryptography and steganography techniques to enhance the security of information. To begin, the secret data was encrypted via enhanced version of AES algorithm. Followed by that RSA algorithm was utilized to covert the encrypted data. Accordingly, this hybrid based method presented two levels of protection. In addition, this hybrid approach has the ability to generate maximum quality stego images and also maximum embedding capability.

Study of Secure Data Transmission-Based Wavelets Using…

217

3 Approaches Used by Existing Researchers In this section we are discussing about the different methods used for secure data transmission among transmitter and receiver using cryptography and steganography techniques. Chandra et al. [4] introduced novel algorithm by which the covert data inside cover image or audio to hide the reality of cipher text and also that image is converted from transmitter to receiver by raising the linking between them to attain data accuracy. Here, RSA cryptographic algorithm, image steganography and RMI framework were utilized for secure data transmission.

3.1 RSA Algorithm The algorithm which utilizes two algebraic structures of public key cryptography method namely Public key as

R = Zn + X

(1)

G = Z (ϕ(n)) · X

(2)

and Public key as

In this algorithm, initially two prime numbers are chosen and their products are utilized to produce both public and private key are provided in Eqs. (1) and (2) respectively. Image steganography–In this process, the images are utilized as covering object of data/information during data communication among sender and receiver. RMI architecture–Remote Method Invocation makes the client–server association. First the server builds an object and also creates its permission slightly. Then the client accepts acknowledge or object reference on the server side and then call upon the function on it. The border utilized by sender-receiver objects to make the inaccessible link is presented via outer layer, stubs, inaccessible reference and also TLP layer. Bandela et al. [5] proposed novel technique comprises of both steganography and cryptography for transmitting data as covert message very securely named as cryptostego technique. This approach helps to locate an image inside another different image. The sendersends the message with encryption to secure the data by hiding with images while the receiver receives the data and perform decryption technique to obtain original image. The proposed framework of data transmission using cryptostego technique is depicted in Fig. 1. Shenoy et al. [6] introduced that the data is hidden by means of Least Significant Bit with images and also the plain text is encrypted using RSA algorithm, the URL is sent to the receiver side with the combination of both cryptography and steganography

218

K. Ravindra Reddy and P. Vijayalakshmi

Fig. 1 Framework for data transmission using crypto-stego

technique by Singh et al. [7] for secure data transmission. The receiver decrypts the message via similar cryptography-based RSA algorithm. If some unknown person tries to steal the data, he can get only URL not the secret data. Kannadhasan et al. [8] focused on merging of cryptography and steganography to transmit the data in a protected manner using Internet of Things. Saraireh et al. [9] proposed filter bank cipher technique which helps to encrypt the text or messages that afforded maximum level of scalability, velocity and protection. The diagrammatic representation of both encryption and decryption process demonstrates by in Fig. 2. The hidden message is encrypted using Advanced Encryption Standard method in sender side and the secret message is retrieved by receiver using steganography based algorithm by Saleh et al. [9] focused on hybrid approach which enhances the data in secure manner during data communication. This hybrid method generates maximum embedding capacity and also quality of stego images increased. Vishnu Babu et al. [10] reviewed many articles regarding secure data transmission between sender and receiver in several methods with cryptography approaches like RSA, AES and steganography algorithms such as LSB, DCT, DWT.

Fig. 2 Diagrammatic representation while data communication by [9]

Study of Secure Data Transmission-Based Wavelets Using…

219

4 Steganography Based Data Transmission Ambika et al. [11] introduced novel approach which has two phases namely data transmission and data extraction. During data transmission, the covert information is preprocessed and encrypted using RSA algorithm in the meantime pre-processing of wrapped image is executed and also knight tour approach is used to create path for knight. During data extraction phase, the information received from sender are decrypted to attain the hidden data using Discrete Wavelet Transform approach which results in secure data transmission. Also, efficient image based data or information linked approach were developed. The essential well designed picture of simple steganography is characterized in Fig. 3. Goudar et al. [12] determines the new technique for safe data broadcasting between transmitter–receiver communication by covering the data with Transmission Control Protocol or Internet Protocol via steganography based approaches. The overall framework for performing encryption and decryption process for secure data transmission is depicted in Fig. 4. Nyo et al. [13] introduced new encryption algorithm especially Twisted exchange method to strengthen the security purpose during data communication among transmitter and receiver. The action performed on sender side is depicted in Fig. 5 and receiver side is shown in Fig. 6.

Fig. 3 Simple steganography by [11]

220

Fig. 4 Overall framework for secure data transmission

Fig. 5 Sender side process while transmitting data

K. Ravindra Reddy and P. Vijayalakshmi

Study of Secure Data Transmission-Based Wavelets Using…

221

Fig. 6 Receiver side process while transmitting data

5 Data Transfer Using Video Steganography For secure data transmission, the sender sends data or message in video format which is explained by Gunjan et al. [14]. This work focused on the combination of both steganography and cryptography algorithm for protecting the data within several multimediafiles such as video, audio and images depicted in Fig. 7. Kolakalur et al. [15] focused on data security by means of hiding data with video series contained by another video series. At sender side, the covert data is concealed with least significant wavelet form in the structure of video format. The hidden data inside video file is retrieved in reverse order of LSB shown in Fig. 8. The metrics such as Peak Signal Noise Ratio, and Mean square error were estimated to find the video series quality. Fig. 7 Encryption and decryption process during transmission

222

K. Ravindra Reddy and P. Vijayalakshmi

Fig. 8 Sending and retrieving digital data using wavelet

In embedding process, the data encryption has done with discrete wavelet transform whereas data decryption has performed using Inverse Discrete Wavelet Transform to retrieve the digital data in stego format.

6 Image Steganography for Data Transmission Saravanan et al. [16] introduced new method for hiding image information by translating it into a different format, which minimizes computing complexity and the proposed diagram is shown in Fig. 9. The data transfer between sender and receiver with high security using DWT (Discrete Wavelet Transform) which has large capacity steganography by Ravi et al. [17]. On the cover picture, Haar and biorthogonal DWT are utilized separately, also on the payload image, Advanced Encryption Standard (AES) along with alterations is utilized to translate the payload image into an encrypted image. The payload image’s coefficients are placed within the cover image’s high frequency bands. Finally DWT algorithm was implemented using ARM8 processor.

7 Data Transmission Using Image Cryptography Murugan et al. [18] explained that discrete wavelet transform is better for secure data transmission when compared to discrete cosine transform approach. This is due to scalability of quality, interest in region coding, low bit rate transmission that is quick to operate, and compatibility with the human visual system, which gives good perception quality. This method extends the merged region and enhances the security. Therefore the outcome of discrete wavelet transform reaches more indiscernible with PSNR ranges from 30 to 54 decibels. This work performed two phases for secure data transmission namely embedding and decoding.

Study of Secure Data Transmission-Based Wavelets Using…

223

Fig. 9 Proposed diagram

Embedding process–The data are hiding secretly with images by means of 2 dimensional Discrete Wavelet Transform and the concealed data are isolated using Haar wavelet. Decoding process–Decoding is similar to decryption process to decrypt the information as plain text for attaining secret message by using two-dimensional Discrete Wavelet Transform. The qualities of images are enhanced via the wavelet coefficient with low level frequency. LPF represents Low Pass Filter which helps to down sample the rows and HPF points the high pass filter for down sample columns as shown in Fig. 10. Kuri et al. [19] proposed IoT based technique along with combination of steganography and cryptography approaches for secure and consistent data exchanging. Moreover, elliptic Galois cryptography was established for encrypting the plain text message and then XOR rules was applied to insert the encrypted data and finally decrypts the message with CNN method to attain original message. Asymmetric key encryption approach is otherwise called Elliptic Galois Cryptography which comprises of CNN encryption and elliptic curve described in Fig. 11. This proposed scheme enhances the effectiveness of estimation and also diminishes the difficulty of rounding errors.

224

K. Ravindra Reddy and P. Vijayalakshmi

Fig. 10 Two dimensional discrete wavelet transform Fig. 11 Elliptic Galois cryptography to perform encryption and decryption

8 Data Transmission Using Cryptography Dogra et al. [20] surveyed many papers for understanding how secure the data is transmitting between sender and receiver in wireless sensor networks. This survey undergoes two phases namely finding and improvement. Finding provides the clarification for competence and dependable protection needs cryptography algorithms. Moreover, this review clarifies the overall security problems in WSN. Chavan et al. [21] suggested symmetric and asymmetric algorithms for secure data communication between sender and receiver. The sender messages are encrypted using private key based algorithm namely AES and the receiver messages are

Study of Secure Data Transmission-Based Wavelets Using…

225

Fig. 12 System architecture of data transmission using RSA algorithm

decrypted using RSA algorithm for enhancing the data security during communication in network. The system architecture of data transmission using RSA algorithm is depicted in Fig. 12.

9 Error Metrics a. PSNR Peak Signal Noise Ratio is the measurement of image quality as a result of embedding which is evaluated in the form of decibels (DB) P S N R = 10log

L2 MES

MES represents Mean Error Square and L point out maximum level. b. MSE Mean square Error defined as to calculate the divergence among original text (plain text) image and stego (cipher text) image. To find the image quality, MSE error metrics are evaluate using the formula

226

K. Ravindra Reddy and P. Vijayalakshmi

MSE =

1 M N (S X Y − C X Y ) X =1 Y =1 MN

Here X and Y represents image coordinates, M points the number of rows, N represents the number of columns, CXY refers the cover image, and SXY refers the stego image. Number of columns, CXY refers the cover image, and SXY refers the stego image.

10 Conclusions We surveyed several articles related with how the cryptographic and steganographic algorithms used during data communication between sender and receiver. For secure data transmission, the sender encrypts the data by hiding with images using cryptographic algorithms and the receiver decrypt the hidden data via decryption process. From this we conclude that these approaches were applied to enhance the data security while performing data transmission in network environment.

References 1. Vasanth, S., & Dhikhi, T. (2016) Secure data transmission using steganography and encryption techniques. International Journal of Pharmacy and Technology, 8(4), 21130–21139. 2. Laskar, S. A., & Hemachandran, K. (2012). Secure data transmission using steganography and encryption technique. International Journal of Cooperative Information Systems, 2, 161–172. 3. Marwa, E., Ali, A. A., & Fatma, A. (2016) Data security using cryptography and steganography techniques. International Journal of Advanced Computer Science and Applications, 7(6). https://doi.org/10.14569/IJACSA.2016.070651. 4. Chandra, S., & Paira, S. (2019). Secure transmission of data using image steganography. ICTACT Journal on Image and Video Processing, August 2019, 10(01). https://doi.org/10. 21917/ijivp.2019.0291. 5. Prasad, P. S., Sunitha Devi, B., Janga Reddy, M., & Gunjan, V. K. (2019). A survey of fingerprint recognition systems and their applications. In A. Kumar, S. Mozar, (Eds.), ICCCE 2018. Lecture Notes in Electrical Engineering (Vol. 500). Springer. https://doi.org/10.1007/978-981-13-02121_53. 6. Manjula Shenoy, K., & Shaikh, S. G. (2019). An approach to secure data transmission through the use of cryptography and steganography. In 2019 International Conference on Communication and Electronics Systems (ICCES), 2019 (pp. 1039–1043). https://doi.org/10.1109/ICC ES45898.2019.9002029. 7. Singh, J., & Sodhi, J. S. (2013). Secure data transmission using encrypted secret message. International Journal of Computer Science and Information Technologies, 4(3), 522–525. 8. Kannadhasan, S., & Nagarajan, R. (2021). Secure framework data security using cryptography and steganography in internet of things. In Multidisciplinary approach to modern digital steganography (pp. 258–279). https://doi.org/10.4018/978-1-7998-7160-6.ch012. (January 2021, book). 9. Fernandes, J. B., Narayan, V., Sammilitha, P. K., Koundinya, P. S., & Krishna, R. R. (2022). Blockchain-based privacy securing G-cloud framework for E-healthcare service. In V. Garcia

Study of Secure Data Transmission-Based Wavelets Using…

10.

11. 12.

13.

14.

15. 16.

17.

18.

19.

20.

21.

227

Diaz, G. J. Rincón Aponte (Eds.), Confidential computing. advanced technologies and societal change. Springer. https://doi.org/10.1007/978-981-19-3045-4_8. Babu, V. S., & Helen, K. J. (2015). A study on combined cryptography and steganography. International Journal of Research Studies in Computer Science and Engineering (IJRSCSE), 2(5), 45–49. Kavitha, A., et al. (2022). Security in IoT mesh networks based on trust similarity. IEEE Access, 10, 121712–121724. https://doi.org/10.1109/ACCESS.2022.3220678 Yellamma, P., Santosh, U. S., Yashwanth, R., Amartya Sai, T. U., & Sampath Kumar, G. N. S. U. (2022). Data security in cloud with hybrid homomorphic encryption technique using GM–RSA algorithm. In V. Garcia Diaz, G. J. Rincón Aponte, (Eds.), Confidential computing. Advanced Technologies and Societal Change. Springer. https://doi.org/10.1007/978-981-193045-4_13. Nyo H. L, Hlaing, A. S., & Maw, T. W. (2018). Secure data transmission using steganography and twisted exchange algorithm. In 2018 IEEE International Conference on Information Communication and Signal Processing (ICICSP), 2018, (pp. 17–21). https://doi.org/10.1109/ ICICSP.2018.8549798. Gunjan, V. K., Kumar, A., & Rao, A. A. (2014). Present & future paradigms of cyber crime & security majors-growth & rising trends. In 2014 4th International Conference on Artificial Intelligence with Applications in Engineering and Technology, Kota Kinabalu, Malaysia, 2014 (pp. 89–94). https://doi.org/10.1109/ICAIET.2014.24. Kolakalur, A., Kagalidis, I., & Vuksanovic, B. (2016). Wavelet based color video steganography. IACSIT International Journal of Engineering and Technology, 8(3) (2016). Saravanan, M., & Priya, A. (2019). An algorithm for security enhancement in image transmission using steganography. Journal of the Institute of Electronics and Computer, 1, 1–8. https:// doi.org/10.33969/JIEC.2019.11001. Lakshmanna, K., Shaik, F., Gunjan, V. K., Singh, N., Kumar, G., & Shafi, R. M. (2022). Perimeter degree technique for the reduction of routing congestion during placement in physical design of VLSI circuits. Complexity, 2022, 11. https://doi.org/10.1155/2022/8658770. Prathima, C., Muppalaneni, N. B., & Kharade, K. G. (2022). Deduplication of IoT data in cloud storage. In C. Satyanarayana, X. Z. Gao, C. Y. Ting, & N. B. Muppalaneni (Eds.), Machine learning and internet of things for societal issues. Advanced Technologies and Societal Change. Springer. https://doi.org/10.1007/978-981-16-5090-1_13. Lakshmanna, K., Shaik, F., Gunjan, V. K., Singh, N., Kumar, G., & Mahammad Shafi, R. (2022). Perimeter degree technique for the reduction of routing congestion during placement in physical design of VLSI circuits. Complexity, 2022(Article ID 8658770), pp. 11. https://doi. org/10.1155/2022/8658770. Dogra, H., & Kohli, J. (2016). Secure data transmission using cryptography techniques in wireless sensor networks: A survey. Indian Journal of Science and Technology, 9(47). https:// doi.org/10.17485/ijst/2016/v9i47/106883. Chavan, A., Jadhav, A., Kumbhar, S., & Joshi, I. (2019). Data transmission using RSA algorithm. International Research Journal of Engineering and Technology (IRJET), 6(3), 34–36.

A Review: Object Detection and Classification Using Side Scan Sonar Images via Deep Learning Techniques K. Sivachandra and R. Kumudham

1 Introduction Many side scan sonar (SSS) systems have been used because they can have image vast areas of the seafloor with a reasonably high resolution in a limited amount of time. Frequently utilized chart in seabed with a high coverage rate for a variety of applications, including development of underwater identification and classification, maritime maps, features of bathymetry and artifacts. Modern UUVs utilize Side Scan Sonar (SSS) as their primary imaging sensor for detecting and classifying (mine-like) objects on the seafloor. These systems are very complicated. It is low-cost, wellunderstood, and widely accessible. However, achieving high resolution and coverage in a large region, it’s impossible to use a traditional SSS and a conventional SSS at the same time. This is a fundamental flaw in these systems. To know about these kinds of information related with underwater acoustics, this paper focused on survey on several existing work. Nayak et al. [18] focused on enhancing the image processing procedure to categorize the probable regions of awareness more efficiently. This work was based on three assistance in the area of underwater robotics comprises of • A pipeline that processes an algorithm for identifying archaeological sites. SSS data was obtained by AUV in order to locate underwater archaeological sites of interest.

K. Sivachandra (B) · R. Kumudham Department of ECE, Vels Institute of Science, Technology and Advanced Studies, Pallavaram, Chennai, India e-mail: [email protected] R. Kumudham e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 V. K. Gunjan et al. (eds.), Modern Approaches in Machine Learning and Cognitive Science: A Walkthrough, Studies in Computational Intelligence 1117, https://doi.org/10.1007/978-3-031-43009-1_20

229

230

K. Sivachandra and R. Kumudham

• Due to the incorporation of traditional feature extraction methods, demonstrated improvements in performance over standard CNN approaches. • A series of AUV deployments in Malta provided experimental validation, resulting in real site identifications.

2 Object Detection in Underwater 2.1 SSS Images Chang et al. [3] tried Fuzzy C-mean clustering method to segment every pixel in image especially SSS image into several clusters that creates shaded portions be partitioned. Here the work focused on solitary simplified underwater object detection method by means of shadow removal of SSS images. This scheme attained better performance in object identification with sturdiness and efficiency. Fuzzy C-means Algorithm To attain shadow segmentation from sonar images, initially they performed image pre-processing to extract the relevant features followed by that applied FCM algorithm which is a repetitive clustering procedure. The least cost function generated to yield an optimal c segment through clustering process was specified as following Eq. (1) JFCM (u, v) =

n  c 

m 2 uik d (xk , vi )

(1)

k=1 i=1

K refers to the pixels in image, I represents cluster, m represents the clustering process quantity, vi represents center of cluster, d denotes the distance between cluster and data. The membership function as well as cluster center are distinct as Eq. 2. uik =  c j=1

n m uik xk , vi = k=1 n m dik 2/(m−1) k=1 uik ( ) 1

(2)

djk

Fuzzy approach comes together to the clustering clarification by means of exploring for least cost function. The FCM algorithm halts while the membership function in some succeeding replications does not alter. The objects in underwater images were distinguished by marking every pixel as the gray value sets to 255 to carry out reintegration and attain customized image named as Y. The object detection and segmentation using underwater images comprise of several steps proceeded by [3]. Initially images were collected regarding sonar, followed by that segmented the images, removal of shadow, and finally detection of objects (Fig. 1).

A Review: Object Detection and Classification Using Side Scan Sonar …

231

Fig. 1 Target Detection and segmentation by [3]

Song et al. [30] focused on segmenting underwater SSS images based on MRF and ELM. Here, the sonar images were partition into several segments for high lightening objects, shading objects, identifying noise regions from SSS images. Initially, the author utilized ensemble method comprised of SE-ELM, ELM-MRF and then SE-ELM-MRF to attain the prediction form. Then the investigation regarding object detection were tested with machine learning algorithms namely xtreme learning, xtreme learning with kernel based, SVM and CNN. Among these algorithms, ensemble technique outperformed in segmentation of sonar images compared to conventional approaches. The flowchart for ensemble algorithm SE-ELM-MRF method is shown in Fig. 2 by [30]. The deep learning based CNN approach were utilized for segmenting sonar images to detect objects in both underwater and surface of the sea. The CNN architecture utilized by [30] is depicted in Fig. 3. Matsuda et al. [16] introduced novel technique known as One Shot Detector that realized object identification exclusive of machine learning approaches and algorithms. OSD is the process of amalgamation of region extraction procedure and outline image procedure. Moreover, the author found that this OSD approach was Fig. 2 Flowchart for ensemble method by [30]

232

K. Sivachandra and R. Kumudham

Fig. 3 CNN framework proposed by [30]

Fig. 4 Segmentation of underwater objects via layers by [16]

completely appropriate for discovering underwater coercion, objects available in sea using SSS images even without using any machine learning technique. The segmentation of images had performed in underwater for identifying threats shown in Fig. 4 by Matsuda et al. The overview of One shot Detector such as image collection, segmentation, shape representation and finally shape matching is illustrated in Fig. 5 by [16]. To improve the overall model “One Shot Detector” performance, the author evaluated classification accuracy which attained around 90% without using any machine learning models. In addition, confusion matrix were estimated for comparing the actual class and predicted class which helps to finding threats in underwater via SSS images easily shown in Table 1. The processing time duration got reduced using region selective sparse coding based SR device which highlights the area of objects in sonar SSS images found by Park et al. [25]. The highlighted area which comprised of identified objects in SSS underwater images hence the successive sparse coding based SR procedure was applied selectively. The efficiency of this work was processing time reduction for reconstructing images nevertheless maintain the similar intensity of perception quality of images as conservative approach. The following Fig. 6 illustrate that the SSS images were generated along with excellent resolution.

A Review: Object Detection and Classification Using Side Scan Sonar …

233

Fig. 5 Overview of object detection and segmentation given by [16]

Table 1 Confusion Matrix done by [16] to distinguish actual and predicted value

Positive

Negative

Positive

89

11

Negative

8

92

Actual class

Predicted class

Fig. 6 Object classification with super resolution by [25]

234

K. Sivachandra and R. Kumudham

Fig. 7 Object classification process proposed by [25]

The image depicted in Fig. 7 how the objects were categorized on underwater SSS images using classification process. Buscombe et al. [5] focused on elimination of acoustic shadows, sea water floor detection, interface for conditions such as while the sea water is excessively muddy or turmoil for consistent intensity echo sounding also seafloor substrate categorization based on single-beam full waveform analysis. In addition these approaches were programmed in an open source software package that make possible to utilize in reconstruction of SSS images automatically. The sonar images were segmented to detect the target objects in underwater through Xtreme machine learning by Song et al. [36]. The author has chosen ELM technique to attain greater performance along with higher learning speed. ELM approach were utilized as classifiers in deep learning based CNN to classify the sonar images in several pathways which can find out mutually narrow and worldwide attributes from SSS images.An effective approach namely subtractive sequential analysis was realized so as to improve the MBES object identification procedure by Brissette et al. [2]. Additionally, automated identification algorithm was utilized effectively on sequentially distinguished images. The comparison was made among the related work, implementation and outcomes of MBES to attain detection accuracy via SSS images. Huo et al. [8] proposed deep learning based CNN algorithm for several classification such as drown, wreck, mine, airplane and seabed in SSS images. At first, the author created real SSS image dataset called Seabed objects-KLSG comprised of wreck 385, drown 36, airplane 62, mine 129 and seabed images as 578. Here, they applied deep CNN with semisynthetic data to train the data that achieved greater accuracy as 97.7% in underwater object detection and classification. The CNN architecture given in Fig. 8 by Huo et al. [8]. Zhao et al. [38] to attain comprehensive seafloor façade image the author introduced SSS image assortment approach via CFP along with restriction of pathway line location. Within the overlap regions, the feature point recognition in sonar image furthermore listing operations was implemented for mutually strips. In accordance

A Review: Object Detection and Classification Using Side Scan Sonar …

235

Fig. 8 Deep based CNN architecture by [8]

with discovered CFP along with pathway line location, fine tuning method was recognized to hold difficult local deformation and also guarantee the worldwide constancy. This process resolves the issue of object interruption as well as no accumulated inaccuracy occurs in mosaicking procedures. The side scan sonar images were segmented using deep based CNN technique to attain greater performance by Song et al. [35].

2.1.1

Automatic AUV

The investigator discovered a scheme for intellectual search of aquaplane through AUV. Here the author applied machine learning methods to recognize the prospective prehistorically sites from AUV on side scan sonar data. Data expansion and image attribute removal are two methods offered in the proposed algorithm to boost performance over conventional CNN approaches. In addition investigational verification in the form of AUV deployments off the coast of Malta, this culminated in the discovery of archaeologically important sites. The block diagram and CNN architecture for side scan sonar processing pipeline is shown in Figs. 9 and 10 by [18]. Gebhardt et al. [4] utilized deep learning based neural network algorithms to resolve the issues of identifying objects such as mine, rock, sand, clay etc. on underwater seabed using side scan sonar images. Moreover, they applied machine learning algorithm namely support vector machine to detect the underwater objects and found the detection accuracy. Also they performed comparison among both machine learning and deep learning algorithm in which deep learning achieved greater accuracy as 93% whereas svm attained 78% in object recognition on seabed. Finally, they

Fig. 9 Block diagram by [18]

236

K. Sivachandra and R. Kumudham

Fig. 10 CNN framework by [18]

insisted that this work were appropriate for embedded utilize with UUV. Li et al. [13] concepts mainly focused on how Deep learning based NN distinguish the objects in SSS images with no training datas/ samples. The proposed deep learning architecture during training phase and also testing phase is described in Fig. 11. Petrich et al. [26] introduced AUVs for improvement and investigational validation of SSS based on self localization algorithms. The relevant attributes with dissimilar classes were extorted from dual frequency sonar images via machine visualization and consequently used to produce sparse bathymetric maps. The range inside the attribute set means attributes of dissimilar sizes and attributes mined at unlike sonar operating frequencies permits for vigorous multi-modal attribute matching. The segmentation of SSS images were executed by AUV for achieving more effective, efficient and exact object identification by Yu et al. [37]. With the aim of better detection of seafloor information the author applied one algorithm which consists of data augmentation, super resolution technology, deep learning and MRF.

Fig. 11 Deep learning architecture by [13]

A Review: Object Detection and Classification Using Side Scan Sonar …

2.1.2

237

Live Recorded Images/Videos

Han et al. [9] the primary key approach for underwater development was to identify and place the foremost object from underwater visualization. To detect the target objects, the author applied deep learning based CNN where the kind of information gathered from Remote Operated Vehicle to attain good results. Moreover, the relevant features mined from the videos/images were carried out using RPN called Region Proposal Network. The author proved that this discovery helpful in real time application for detecting objects via videos in seabed or underwater acoustics. The deep learning based CNN architecture were proposed by [9] for extracting features from images, object recognition and classifying the objects from underwater images is depicted in Fig. 12. Nikolovska et al. [21] object recognition enhancement attained by fusing information from specific sensors namely Recorded camera, SSS, BOSS and MBES. The main goal was time interruption beam forming which generated the outcomes as Voxel layer as well as synthetic Aperture Sonar processing. However, the recognition of objects and classification of objects as mine or non-mine from sensor based sonar images had performed with greater precision within least time duration. Through remote sensing approach, the objects found on the ocean floor were identified using SSS images established by Okoli et al. [24]. Similar to object detection, the documents were recognized using gradient based learning algorithm proposed by Lecun et al. [34]. Generative Adversarial Network based algorithm was introduced by Sung et al. [32] which produce realistic sonar images. This deep learning based approach were utilized for target recognition which performed the underwater activities like wreckage, unwanted feature elimination, as well as landmark based navigation were computerized. Fig. 12 Object recognition and classification by layers in CNN [9]

238

2.1.3

K. Sivachandra and R. Kumudham

Sonar Images

Ochal et al. [23] introduced deep learning based CNN normally applied for object detection in underwater acoustics on both images (Sonar, optical). Moreover, huge amount of data or images can be moderately very expensive and prolonged too while examining extraordinary objects or executing real time processes. The author discovered novel approach namely Few Shot Learning were utilized for object recognition in underwater seabed automatically as well as enlarged their knowledge competence. Comparison had prepared among supervised and semi-supervised using underwater optical and SSS images. Using FSL methods, an average gain of 7% accuracy points over the ConvNet baseline, with up to 8.6% on optical datasets while on sonar images, up to 16.8%. Anitha et al. [1] applied image processing technique for recognition of objects either on seabed or under the seawater using sonar images. The input images given by sonar technology were gray scale images forever. The step by step procedure was put forth by [1] discovery of objects in sonar images such as removal of noises from images, detection of edges, and finally image categorization. Among these three steps, detection of edges in images plays a significant role in image discovery and classification. Here, assessments were performed among edge based detection via fuzzy approach and block processing approach. Lee et al. [28] introduced a real time simulated dataset for object recognition in underwater sonar images using style transferred learning technique. To train the images related with sonar, style transfer learning methods were applied for object identification in underwater acoustics. Feldens et al. [7] determined network based structure specifically RetinaNet for object recognition in underwater via sonar images in addition to distinguish rocks from mine in backscatter mosaics rooted from SSS working at 384 kHz. The object detection and categorization depended straightforwardly on accurate resolution of underwater acoustics images were performed involuntarily by Langner et al. [12]. The author collected images from FU-Berlin and also FGANFOM for detecting objects and undergoes classification too. The procedures followed by [12] are Pre-processing the images comprised of normalization, altitude evaluation along with slope range improvement, and finally geo-referencing. Several screening approaches like arithmetical attributes, threshold segmentation, analysis of shadow region in sonar images, one dimensional cross association, MSER method, k means algorithm, highly order statistic based segmentation were applied for implementing sonar images for object recognition and classification. Probabilistic based NN were utilized for classifying images for distinguishing mine or rock. The datasets were gathered lively by SeaOtter AUV from Atlas Elekronik, Batli and Mediterranean Sea in which dissimilar SSS schemes were utilized in different zones. These algorithms were evaluated by monitoring the sea approximately 25 km2 of the seabed. As a result of negative cause of sound on several object recognition, Nguyen et al. [20] applied clustering approaches such as k-means, density based clustering equally on sonar based images and 3-D point cloud LiDAR datas to enhance the performance in underwater object detection. Especially, the density based clustering with noise suitable to remove noises in sonar images. The author proposed this method in domain

A Review: Object Detection and Classification Using Side Scan Sonar …

239

namely maritime application regarding object recognition, and autonomous driving system. The CNN structure designed in Fig. 13 by Nguyen et al. [20] as a overall framework for object recognition, object classification using underwater SSS images. The Fully Convolution Network model formation in Fig. 14 was designed by [20] to extort multiple objects out of the background in underwater sonar images to organize the input data for the clustering methods. Priyadharshini et al. [27] introduced edge based segmentation approach which utilized the morphological operations for finding edges in sonar images followed by that they proposed object tracing algorithm to track the objects in underwater acoustics on sonar images. The sonar images were pre-processed using filtering techniques such as median filter, Wiener filter and also morphological gradient were attained by means of deduct dilated and eroded image. Sheffer et al. [29] determined the procedure for modifying as well as restructuring [11] the image based on sonar that uses normalization comprised of slant range

Fig. 13 CNN architecture by [20]

Fig. 14 FCN for segmenting underwater objects by [20]

240

K. Sivachandra and R. Kumudham

Fig. 15 Reconstruction of original image via AUV sensors [29]

correction, Yaw and pitch correction, and speed and location correction shown in Fig. 15. This procedure were successfully completed via navigation and inertial data attained through AUV sensors. Four (A, B, C, D) autonomous steps which had done in analogous to reconstruct the original images as corrected one. Doherty et al. [6] introduced three phases for object identification and classification were performed as Phase 1: Object recognition-utilizes thresholding approach attached with adaptive averaging approach to place OOI in SSS images. Phase 2: Classification-applied binary classification algorithm for classifying objects as either mine or non-mine. This categorization was attained with feature based decision tree. Phase 3: Identification-Mark the regions where the objects found in sonar images. Sung et al. [31] identified the cross talk noises on sonar images through image based neural network framework and removed noises from that specified images based on detection identification. To attain their target and improve the consistency of proposed approach, the author applied NN method to 3-D point cloud creation and produced more number of precise cloud point. Sung et al. [32] introduced deep learning approach for underwater object detection that is still needed to automate underwater activities like wreck examination, extract reduction, and landmark based navigation. This approach manufactured real sonar images using GAN as Generative Adversarial Network. This paper proposed methods that synthesize realistic sonar images using a Generative Adversarial Network.

2.1.4

Hardware Based Object Detection

Hossain et al. [10] introduced deep learning technique for identifying trashes on underwater sonar images which helps in object detection and classification. In this, they attached ultrasonic sonar sensor on robot which identifies the objects besides the pathway. Also, the camcorder unit launches the images of sonar objects to the specific hardware mentioned as Raspberry pi to categorize the images as either trash or non-trash. The robot utilized for implementation discover broad assortment of trash that reaches very accuracy and least expenditure too. The hardware utilized for implementation are Arduino Mega 2560, Ultrasonic Sensor (HC-SR04), Motor Driver (L298N), Raspberry Pi 3 Raspberry Pi camera module, DC Motors, Servo Motors (MG996R), and Power source. The CNN framework in Fig. 16 is illustrated. The modules for trash discovery proposed by [10] are

A Review: Object Detection and Classification Using Side Scan Sonar …

241

Fig. 16 CNN layers for target detection by [10]

Detection of objects on sonar images-Here the objects that are less that 30 cm were detected using ultrasonic sensor, then those objects are moved to hardware (Raspberry pi) and trashes had found using deep learning technique. After the discovery of objects, it was to be recognized as trash or non trash. Initially the image resolution was specified as 64*64 pixels. Then data were divided into 70%, 20% and 10% as for training, testing and validation phases. Followed by that, they applied Convolutional Neural Network layer of 1D, 2D and 3D for attaining better optimization. The overall CNN layers used for detecting trashes are depicted in Fig. 16. Once the objects were identified as trashes using deep learning technique, then Raspberry pi drive the alarm to Arduino which made started the servo motor to function.

2.1.5

Bubble Detection

Several investigators experimented based on object detection in underwater, achieves greatest outcomes along with higher performance. Analogous to object detection in underwater sea/ocean, or surface of the sea/ocean, Uchimoto et al. [33] detected bubbles in shallow sea or ocean floor on SSS images that makes helpful for maritime applications in real time.

3 Pipeline Detection in Underwater Prasad et al. [14] introduced that the pipeline detection in underwater sea for investigation from visual data incorporated with pictures or videos which were analyzed via machine learning along with deep learning algorithms. Hence the pipelines which were collected from deep sea videos recorded from the attached camera at LVOONL islands located in Norway which was detected at the deepness of 260 m to attain better

242 Table 2 Metrics used along with formula by [14]

K. Sivachandra and R. Kumudham

Metrics used

Related formula

Accuracy

Number of truely predictions Total number of observations/samples FP FP+TN TP TP+FN

FPR or Sensitivity TPR or Specificity

Fig. 17 Object classification proposed by [14]

(a)

(c)

(b)

(d)

(e)

(f)

(g)

(i)

(k)

(h)

(j)

(l)

(m)

performance in detection as well as categorizing outcomes with accuracy measure as 76.2% and AUC as 87.5%. Here, every classifier was validated by tenfold cross validation by examining that the essentials of every class were spread equally in every one of the folds. The performance of the models was evaluated by certain metrics namely accuracy, loss, and area under the curve (AUC) average scores (Table 2). Finally the confusion matrix was evaluated to compare the actual value with the predicted value for classifying the images or videos into mine like objects in underwater seabed. The sample figure illustrates that the discovered species found from recorded video utilized for constructing the training dataset for indication to perform classification automatically as depicted in Fig. 17.

4 Other Existing Literature Studies in Underwater Object Detection Wang et al. [22] reviewed several articles for detection of objects on underwater and also sea surface too using deep learning approaches. In that reviewed paper, the author investigated wide concepts of existing work such as what dataset they utilized, architecture used for object identifying and classification using deep learning

A Review: Object Detection and Classification Using Side Scan Sonar …

243

algorithms, along with rigorous comparisons too. Finally, the author reviewed the several modern network layer in deep learning approaches were drawn. SSS sensors are often used by maritime archaeologists when reviewing a wide region in explores of wreckage or further antiquarianism sites of significance. Moniruzzaman et al. [17] explained the importance of deep learning technique for analyzing underwater images within modern ancient times. Diesing et al. [15] the author carried out a stateof-art of widespread apply in global classifying images based on remote sensing data, the relevant authority even though along with a superior methodology as well as past history. Neupane et al. [19] surveyed the latest object identification, tracing, and recognition via deep learning approaches. A complete survey of existing work had done by several researchers, and essential facts concerning the data gathering procedure, images utilized, and information about hyper-parameters were presented.

5 Overall State-of-Art This literature assessment discussed regarding authors, published year, images used, objectives, techniques and algorithms applied detection and classification accuracy for finding performance in object recognition in underwater acoustics, various outcomes achieved by various existing researchers summarize in Table 3.

6 Conclusion In this study, the recent methods and algorithms for recognizing, categorizing several underwater sea/oceanic/aquatic objects using Artificial Intelligence namely machine learning and deep learning approaches were discussed. Based on targets/objects detection in underwater environments, the algorithms were utilized. Moreover, AUV were implemented by several investigators for underwater object detection and distinguishing the objects into either rock or mine or clay. Every authors deep learning based CNN proposed framework were conversed. Also, the published year, objectives of previous work, what are the techniques and algorithms utilized, identification and classification accuracy to analyze the model performance, and outcomes found by the existing investigators were summarized as state-of-art table. In addition, the challenges faced in target detection were highlighted also the application were listed. This reviewed work will applicable for receiving the overall ideas regarding object detection and classification objects in underwater sea/oceans using sonar, SSS images, live recorded videos/images makes clear-cut view for further investigation.

ML

98.5%

SVM + ELM + MRF

Segmenting SSS images

SSS images (9 images)

Song et al. [30]

Detection time 85 ms

The key technique of the DL underwater robot development is to detect and locate the main target from underwater vision

Dataset was prepared using an underwater video obtained from a sea cucumber fishing ROV

Han et al. [9]

86%, –

87.5%, 76.18%

Detection/ classification accuracy

DCNN

Fuzzy C-mean clustering (FCM)

Method of Shadow-Removal in benthic object detection

Sonar

Chang et al. [3]

ML

SVM

Proposed an automatic ML pipeline for underwater animal identification and categorization, performed filtering using machine learning techniques

8818 images

Algorithm

Lopez-Vazquez

Technique used

Objectives

Images

Authors

(continued)

The overview performances of SE-ELM-MRF were verified through the good segmentation results on sonar images of dissimilar places

The detection and classification of marine organisms

Otsu for partitioning of images

Detecting and categorizing elements in underwater via live taken videos and images

Outcomes

Table 3 Overall State-of-art for object detection and classification in underwater by published year, images used, objectives, and techniques applied, accuracy, outcomes

244 K. Sivachandra and R. Kumudham

639 SSS images of 14 categories

Li et al. [13]

Selective coding

SSS underwater images

Park et al. [25]

Objects is detected for side scan sonar based underwater images

Segmentation approach

ML, DL

DL

Technique used

Acoustics images captured via To detect objects in Edgetech 4125 SSS device underwater acoustics

To find multiple object in underwater acoustics

To solve the issues on zero-shot SSS image classification problem

Objectives

Priyadharsini and Sreesharmila [27]

Nguyen et al. [20] SSS images

Images

Authors

Table 3 (continued)

100%, –

–, 83.21%

Detection/ classification accuracy

Region selective sparse coding

96.9%, –

Moore’s Neighbor – algorithm

K-means clustering, DBSCAN

DNN Zero shot object classification method

Algorithm

(continued)

To reduce processing time and also objects in underwater were detected

Detecting the objects in seafloor and performed segmentation of images

To recognize multiple objects in underwater as well as terrestrial surroundings

Enhance excellent classification ability without any training samples

Outcomes

A Review: Object Detection and Classification Using Side Scan Sonar … 245

To fill the segmentation maps according to the class of each pixel. Therefore, peak signal to noise ratio (PSNR) of the fine synthetic images with the Gaussian-filtered SSS images that can be used as one evaluation metric for segmentation performance

SSS images

SeabedObjects-KLSG dataset will be open on Github

Realistic Sonar images

Song et al. [30]

Huo et al.

Sung et al. [32] REALISTIC SIMULATION USING RAY TRACING AND GAN to synthesize realisticsonar images

Multi-class classification on sonar images

Objectives

Images

Authors

Table 3 (continued)

DL

DL, ML

Technique used

Ray-tracing method to estimate semantic information

Deep transfer learning, semi-synthetic training data

Non-parametric algorithm and ELM, CNN

Algorithm

–, 97.76%

Detection/ classification accuracy

Object detection and segmentation

Classifying objects in underwater

PSNR and Gaussian filtered both utilized as one estimation metric for objects segmentation performance

Outcomes

246 K. Sivachandra and R. Kumudham

A Review: Object Detection and Classification Using Side Scan Sonar …

247

References 1. Anitha, U., &Malarkkan, S. (2016). Analysis of edge detection techniques for side scan sonar image using block processing and fuzzy logic methods. Advances in Intelligent Systems and Computing, 363–370. https://doi.org/10.1007/978-981-10-1678-3_35. 2. Brissette, M., & Clarke, J. (1999). Side Scan versus multibeam echosounder object detection: A comparative analysis. International Hydrographic Review, 76. 3. Chang, R., Wang, Y., Hou, J., Qiu, S., Nian, R., He, B., & Lendasse, A. (2016). Underwater object detection with efficient shadow-removal for side scan sonar images. In OCEANS 2016, Shanghai. https://doi.org/10.1109/oceansap.2016.7485696. 4. Kar, S., & Kar, S. (2023). Pedagogical considerations in the new normal: From tradition to technology. In: A. Choudhury, A. Biswas, & S. Chakraborti, (Eds.), Digital learning based education. Advanced Technologies and Societal Change. Springer. https://doi.org/10.1007/ 978-981-19-8967-4_6. 5. Buscombe, D. (2017). Shallow water benthic imaging and substrate characterization using recreational-grade sidescan-sonar. Environmental Modelling & Software, 89, 1–18. 6. Doherty, M. F., Landowski, J. G., Maynard, P. F., Uber, G. T., Fries, D. W., & Maltz, F. H. (n.d.). Side scan sonar object classification algorithms. In Proceedings of the 6th International Symposium on Unmanned Untethered Submersible Technology. https://doi.org/10.1109/uust. 1989.754734. 7. Feldens, P. (2020). Super resolution by deep learning improves boulder detection in side scan sonar backscatter mosaics. Remote Sensing, 12, 2284. 8. Huo, G., Wu, Z., Li, J. (2020). Underwater object classification in sidescan sonar images using deep transfer learning and semisynthetic training data. IEEE Access. https://doi.org/10.1109/ ACCESS.2020.2978880. 9. Han, F., Yao, J., Zhu, H., & Wang, C. (2020). Marine organism detection and classification from underwater vision based on the deep CNN method. Mathematical Problems in Engineering, 2020, 1–11. 10. Shaik, A. S., Karsh, R. K., Suresh, M., & Gunjan, V. K. (2022). LWT-DCT based image hashing for tampering localization via blind geometric correction. In A. Kumar, S. Senatore, & V. K. Gunjan (Eds.), ICDSMLA 2020. Lecture Notes in Electrical Engineering (Vol. 783). Springer. https://doi.org/10.1007/978-981-16-3690-5_156. 11. Jang, J., Shin, M., Lim, S., Park, J., Kim, J., & Paik, J. (2019). Intelligent image-based railway inspection system using deep learning-based object detection and weber contrast-based image comparison. Sensors (Basel, Switzerland), 19. 12. Muthmainnah, Ganguli, S., Al Yakin, A., & Ghofur, A. (2023). An effective investigation on YIPe-learning based for twenty-first century class. In A. Choudhury, A. Biswas, & S. Chakraborti (Eds.), Digital learning based education. Advanced Technologies and Societal Change. Springer. https://doi.org/10.1007/978-981-19-8967-4_2. 13. Li, C., Ye, X., Cao, D., Hou, J., & Yang, H. (2021). Zero shot objects classification method of side scan sonar image based on synthesis of pseudo samples. Applied Acoustics, 173, 107691. https://doi.org/10.1016/j.apacoust.2020.107691 14. Prasad, P. S., Beena Bethel, G. N., Singh, N., Kumar Gunjan, V., Basir, S., Miah, S. (2022). Blockchain-based privacy access control mechanism and collaborative analysis for medical images. Security and Communication Networks, 2022(Article ID 9579611), 7. https://doi.org/ 10.1155/2022/9579611. 15. Diesing, M., Mitchell, P., Stephens, D. (2016). Image-based seabed classification: What can we learn from terrestrial remote sensing? ICES Journal of Marine Science, 73(10), 2425–2441. https://doi.org/10.1093/icesjms/fsw118. 16. Matsuda, Y., Ogawa, M., & Yano, M. (2015). System of detecting underwater threats in side scan sonar images. In OCEANS 2015-Genova. https://doi.org/10.1109/oceans-genova.2015. 7271585.

248

K. Sivachandra and R. Kumudham

17. Moniruzzaman, M., Islam, S. M. S., Bennamoun, M., & Lavery, P. (2017). Deep learning on underwater marine object detection: A survey. Lecture Notes in Computer Science (pp. 150– 160). https://doi.org/10.1007/978-3-319-70353-4_13. 18. Nayak, N., Nara, M., Gambin, T., Wood, Z., & Clark, C. (2019). Machine learning techniques for AUV side scan sonar data feature extraction as applied to intelligent search for underwater archaeological sites. (25 Aug 2019). 19. Ahmed, S. M., Joshitha, D., Swathika, A., Chandana, S., Sahhas, & Gunjan, V. K. (2023). Dietary assessment by food image logging based on food calorie estimation implemented using deep learning. In: A. Kumar, S. Mozar, & J. Haase, (Eds.), Advances in cognitive science and communications. ICCCE 2023. Cognitive Science and Technology. Springer. https://doi. org/10.1007/978-981-19-8086-2_107. 20. Nguyen, H.T., Lee, E., Bae, C.H., & Lee, S. (2020). Multiple object detection based on clustering and deep learning methods. Sensors (Basel, Switzerland), 20. 21. Nayak, S. C., Das, S., Sahoo, B., & Satyanarayana, B. (2023). AEFA-ANN: Artificial electric field algorithm-based artificial neural networks for forecasting crude oil prices. In: M. Usman, X. Z. Gao, (Eds.), Embracing machines and humanity through cognitive computing and IoT. Advanced Technologies and Societal Change. Springer. https://doi.org/10.1007/978-981-194522-9_9. 22. Wang, N., Wang, Y., Er, M. J. (2020). Review on deep learning techniques for marine object recognition: Architectures and algorithms. Control Engineering Practice, 2020, 104458. ISSN 0967–0661, 23. Ochal, M., Vazquez, J., Pétillot, Y., & Wang, S. (2020). A comparison of few-shot learning methods for underwater optical and sonar image classification. http://arxiv.org/abs/2005.04621. 24. Vamseekrishna, A., Siva Ganga Prasad, M., Gopi Krishna, P., Bhargavi, P., Rohit, S., Tanmayi, B. (2023). Low-cost ECG-based heart monitoring system with ubidots platform. In: M. Usman, X. Z. Gao, (Eds.), Embracing machines and humanity through cognitive computing and IoT. Advanced Technologies and Societal Change. Springer. https://doi.org/10.1007/978-981-194522-9_6. 25. Park, J., Ku, B., Jin, Y., & Ko, H. (2019). Side scan sonar image super resolution via regionselective sparse coding. IEICE Transactions on Information and Systems, 102-D, 210–213. 26. Petrich, J., Brown, M. F., Pentzer, J. L., & Sustersic, J. P. (2018). Side scan sonar based selflocalization for small Autonomous Underwater Vehicles. Ocean Engineering, 161, 221–226. https://doi.org/10.1016/j.oceaneng.2018.04.095 27. Priyadharsini, R., Sreesharmila, T. (2019). Object detection in underwater acoustics images using edge based segmentation method. In ICRTAC 2019. 28. Gunjan, V. K., Shaik, F., Venkatesh, C., Amarnath, M. (2017). Visual quality improvement of CT image reconstruction with quantitative measures. In Computational methods in molecular imaging technologies. SpringerBriefs in Applied Sciences and Technology(). Springer. https:// doi.org/10.1007/978-981-10-4636-0_4. 29. Sheffer, T., & Guterman, H. (2018). Geometrical correction of side-scan sonar images. In 2018 IEEE International Conference on the Science of Electrical Engineering in Israel (ICSEE). https://doi.org/10.1109/icsee.2018.8646188. 30. Song, Y., He, B., Liu, P., & Yan, T. (2019). Side scan sonar image segmentation and synthesis based on extreme learning machine. Applied Acoustics, 146, 56–65. https://doi.org/10.1016/j. apacoust.2018.10.031 31. Sung, M., Cho, H., Kim, T., Joe, H., & Yu, S. (2019). Crosstalk removal in forward scan sonar image using deep learning for object detection. IEEE Sensors Journal, 19, 9929–9944. 32. Sung, M., Kim, J., Lee, M., et al. (2020). Realistic sonar image simulation using deep learning for underwater object detection. International Journal of Control, Automation and Systems, 18, 523–534. https://doi.org/10.1007/s12555-019-0691-3 33. Uchimoto, K., Nishimura, M., Watanabe, Y., & Xue, Z. (2019). Bubble detection with side-scan sonar in shallow sea for future application to marine monitoring at offshore CO2 storage sites. 34. LeCun, Y., Bottou, L., Bengio, Y., & Haffner, P. (1998). Gradient-based learning applied to document recognition. Proceedings of the IEEE, 86(11), 2278–2324.

A Review: Object Detection and Classification Using Side Scan Sonar …

249

35. Song, Y., Zhu, Y., Li, G., Feng, C., He, B., & Yan, T. (2017). Side scan sonar segmentation using deep convolutional neural network. In OCEANS 2017-Anchorage, Anchorage, AK, USA, 2017 (pp. 1-4). 36. Kumar, S., Ansari, M. D., Gunjan, V. K., Solanki, V. K. (2020). On classification of BMD images using machine learning (ANN) algorithm. In: A. Kumar, M. Paprzycki, V. Gunjan, (Eds.), ICDSMLA 2019. Lecture Notes in Electrical Engineering (Vol. 601). Springer. https:// doi.org/10.1007/978-981-15-1420-3_165. 37. Yu, F., Zhu, Y., Wang, Q., Li, K., Wu, M., Li, G., Yan, T., He, B. (2019). Segmentation of side scan sonar images on AUV. In 2019 IEEE Underwater Technology (UT). https://doi.org/10. 1109/ut.2019.8734433. 38. Zhao, J., Shang, X., & Zhang, H. (2018). Side-scan sonar image mosaic using couple feature points with constraint of track line positions. Remote Sensing, 10(6), 953. https://doi.org/10. 3390/rs10060953

Analysis of High Performance Optical Networks Using Dense Wavelength-Division Multiplexing Application L. Bharathi, N. Sangeethapriya, J. Prasanth Kumar, and G. Sandeep

1 Introduction Fiber connects and central optical Networks associated with optical exchanging components, from one side to the other, can support a huge measure of traffic for global or even international data transmission and empowerment for data trading and serve many customers [1]. Network communication application requirements have shaped the development of events and configured/programmed optical communication [2]. In addition to the characteristics of programming configured on the control plane and the development of bringing the administrative plane together, naturally configured optical communication needs a keen calculation to advise the control plane communication [3]. The network traffic of nearby network access networks is collected and maintained in communication ring communication, allowing numerous client supporters and centralized optical communication [4]. The electronic edge switches of central optical communication are essentially the total number of traffic streams from communication [5]. Figure 1 gives the: Optical Network (ON) administrators experienced a wide range of transfer speed prerequisites, driven by the knowledge that future communication should offer higher limits, higher adaptability, and higher uninterrupted quality. Besides, to meet these highlights, such optical communication should offer multiplexing exchange functions, which were seen as a financially effective way to reduce general organization speculation and activity costs [6]. Accordingly, the Frequency Division Multiplexed Optical System (FDMOS) is now generally called upon to assist in high data transfer capacity, significant distance transmission [7]. The Optical Network (NW) is based on improving the network process and establishing communication connected to the Internet and internationally [8]. L. Bharathi (B) · N. Sangeethapriya · J. P. Kumar · G. Sandeep Department of ECE, Ramachandra College of Engineering, Eluru, Andra Pradesh, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 V. K. Gunjan et al. (eds.), Modern Approaches in Machine Learning and Cognitive Science: A Walkthrough, Studies in Computational Intelligence 1117, https://doi.org/10.1007/978-3-031-43009-1_21

251

252

L. Bharathi et al.

Channel Coding

Data Symbol

Data Packet

Mapping

Information

(Input)

Pulse shaping Conversion

Output

Modulated

Signal

Carrier Signal

Fig. 1 Block diagram of optical network communication

Adjustments should be made to accomplish certain objectives, for example, load adjusting, green systems administration, limit boost [9]. It is similarly described as a configuration of the draw out organization stage to determine the ideal processes from the structure’s configuration to manage assets’ distribution [10]. Despite the stretched configuration, the transient configuration is more solid than the optical communication configuration [11]. A traffic request is usually generated with a short interval between the configuration phase and the sending phase [12]. Network communication certainty of a real foundation, the inability of a real level when a sub-ideal property or path can be given due to calculation or a specific strategy [13]. This proposal centers around a configured transient optical organization, including directing calculations, frequency function, recovery recovery-based optical organization, traffic prep strategy, and optical light path security [14]. In Optical Network (ON) communication operating in a non-direct system, the tool to allow basic and productive non-line vulnerabilities is fundamental [15]. Such a device should anticipate the nature of the transmission of light paths traveling through different steering lanes in optical communication, which leads to a pattern of non-direct weak interaction between those signals. The direct non-interruption approval device should also capture these various highlights to assess the signal quality and the forecast’s completion [16]. To improve the organization’s limits in this new period of sending rational broadcasts, the restricted organization should make productive use of assets [17]. In any case, what are the assets and inevitability in optical communication and how can use these resources within the limits of the increase in multifaceted information traffic and research into the imaginative development that comes with sending the fifth era of radio access networks became progress to enable more bandwidth.

Analysis of High Performance Optical Networks Using Dense …

253

2 Related Work Optical Network (ON) correspondence is one of the promising answers for developing transfer speed in full-scale networks for a wide range of indoor and outdoor applications. In miniature-sized communication, Optical Network (ON) is indicated at the end as an interconnect innovation to provide effective correspondence in a Network On-Chip (NOC) [18]. The uninterrupted expansion in the thickness of the preparation centers is not only based on individual metal interconnections, as they have a congenital barrier to correspondence data transmission and the use of force [2, 14]. Optical correspondence with thin linewidth laser sources makes the pulse between wanted and moderate icons crossest lock-in communication with simultaneous data transmission on an Electron Detector (ED). However, the effect of the barrier on correspondence has not been investigated; also, the source of information may be noncompressed, as opposed to many distributed functions [3]. Indoor infrared remote correspondence arranged in a Bit Obscure Probability (BEP), where the be beat barrier restriction system is considered. In such communication with large linewidth optical sources (e.g., Light Emitting Diodes (LEDs)), striking commitments become negligible, resulting in an improvement on the barrier [4]. Error analysis cannot be applied with limited linewidth laser sources; inconsistent quality is a prerequisite for correspondence structure, which is regularly estimated in words. Investigation of other correspondence enforcement measures [5, 17]. By configuring a blockchain mindful structure, the organization architect can find the ideal setup that meets the dependencies required for non-chip correspondence. For instance, the number of connections using the same frequency, and with the goal of distance, the distance that completes the required degree of reliability will separate them [6]. The optical encoding scheme is based on ambiguous stage encoding, in which the optical code is randomly selected from the code set for code jumping. Security enforcement is being investigated [7]. Space-Division Multiplexing Elastic Optical Networks (SDM-IONS) will play an important role in trending their expanded Internet traffic due to their range of use adaptability and prevailing limitations [8]. However, except for Personal Physical Level Defects (PLIs), the recently released Cross Stall Luck (CST) confirms the transmission quality in conjunction with the future administration’s flight. As such, planning more clever and powerful asset task calculations can be daunting [9]. The rise of human-made brainpower provides an uncertain answer to such issues. The standard response to short-pull fiber-optic interchanges is not to send a sensible system, i.e., to tweak and only to identify light power [10]. In such a system, the sign interferes with the Optical Network (ON) and is confused with the electrical smoke. The limits of any sensible optical connections have been broadly focused on optical seasons or cozy crunch. It can complete the integral security level with a short-wave example, expertly improving the structure’s property [11, 16].

254

L. Bharathi et al.

The business opportunity, communication and research communication are examining the next steps to take the market [12, 16]. One promising empowerment innovation has emerged in response to the Passive Optical Network (PONs), providing enormous data transmission for correspondence and supporting extended interest [13, 17, 18]. Previously it had become one of the backbones for the giant for the transmission of information. Getting closer to the final client is becoming more common as the need for high-speed data transmission develops [14]. In this configuration of integrating and empowering innovations, the task examines the source for providing different repetition channels. The Fourier transformation varies rapidly from the optical space to the repetition field, and so on, and the use of different regulation systems [15, 18].

3 Materials and Method Fiber optics have become the center of our broadcast communications and information planning system. Second optical fiber is the preferred transmission method for any information over two or three megabytes per second and over one kilometer. The original fiber optic associations were used. The optical fiber is a substitute for copper links for long-distance piece transmission over long distances. Another era of fiber-optic associations is just emerging. Dense Wavelength-Division Multiplexing (DWDM) associations use the fiber limit to complete the normal bandwidths of a few gigabytes per second to terabit every second. Besides, they misuse the direction and exchange of signals in the field. Figure 2 gives the: rapid advancement of innovation, coupled with an inexplicable interest in data transmission, is bringing rapid progress of this communication from laboratories to commercial centers. The Optical Network (ON) fiber transmission is the well-covered perspective of systems such as broadcast vision and design and control and board issues of second-year fiber optic communication. Such a system would not end without showing the expected parts of building this communication, especially since the organization models rely heavily on them. Individuals planning optical communication should be aware of their potential system, therefore, attempts to cover systems administration issues with segments, transmissions and other optical bodies

3.1 Source Information (Input) and Electrical Stage Transmitter The electric sign is applied to the optical transmitter, and transmitters include a driver circuit, a light source, and a fiber fly lead. The driver circuit operates thelight source replaces the electrical sign with the optical sign. Fiber fly lead is used to attach

Analysis of High Performance Optical Networks Using Dense …

Electrical Stage (Transmitter)

Source Information (Input)

255

Dense Wavelength-Division Multiplexing (DWDM) (Prescient Control Algorithms)

Transmission medium as Optical Fiber

Photodetector

Electrical Stage

Destination

(Receiver)

Information (Output)

Fig. 2 Proposed Block diagram for Optical Network

optical signals to optical fiber and focuses on experts who need to find out about network organizers, architects or administrators, graduate undertakings in electrical designing and software engineering, and optical institutes.

3.2 Optical Network Fiber optic connectors and links are available in virtually every interchange project. A base station with a remote back hole ensures that fiber jumpers and cabling are being used somewhere in that organization. Having an overall understanding of fiber optics, what’s more, the specific fiber and connector types that are accessible will allow you to have a more beneficial discussion with client’s more optical fiber, the link, otherwise called the rounded and hollow dielectric waveguide. Optical fibers consider the conditions, elasticity, rigidity and discomfort that work in the same climatic conditions gives the fiber-optic link is made of excellent tempered glass (C) or plastic and is acceptable and model given shown below in Fig. 3.

256

L. Bharathi et al.

Fig. 3 Module of fiber optical cable

3.3 Photodetector Optical fiber is a dielectric waveguide or medium where data (sound, information or video) is sent as light through glass or plastic fibers. The cladding supports the waveguide structure, the center by retaining contaminants to the surface and, when thick enough, generously reduces radiation’s inconvenience in the surrounding air. The main elements of glass are usually less unfortunate in the examination of the central elements of plastic. Moreover, most strands are embedded in a flexible, scraped area safe plastic material that precisely disintegrates the fibers from minor mathematical anomalies and distortions. A set of guided electromagnetic waves, similarly called waveguide methods, can illustrate light’s propagation along a waveguide. Only a specific number of mods are equipped to spread through the waveguide.

3.4 Dense Wavelength-Division Multiplexing (DWDM) Optical fiber is a dielectric waveguide or medium where data (sound, information or video) is sent as light through glass or plastic fibers. It consists of a straight center with a refractive list contained by the direct cladding of an exceptionally low refractive record; the refractive list of cladding is no less than that of the centermost common qualities right now are a center refractive record and a cladding list Dense Wavelength-Division Multiplexing (DWDM) network comprises frequency directional hubs connected by a hidden point. Figure 4 give the : Frequency Division is a successful method of abusing the enormous data transfer capacity of Optical Network (ON) to meet the rapid growth of transfer speed interest on the Internet.

Analysis of High Performance Optical Networks Using Dense …

257

Fig. 4 Architecture of Dense Wavelength-Division Multiplexing (DWDM)

3.4.1

Prescient Control Algorithms

The cladding waveguide lowers the structure, scavenges the center from impurities covering the surface and, when thick enough, generously reduces the misfortune of radiation in the contained air. The main elements of glass are generally less unfortunate than the central elements of plastic. Moreover, most of the fibers are exemplified inflexible, scraped, safe plastic material that separates the strands from slight mathematical distortions and deformities. An algorithm technique is a set of guided electromagnetic waves, also called waveguide methods, that can illustrate light’s propagation along a waveguide. Step 1: Initialize the input data. Step 2: Identify the basic signal of fiber optic communication. Ir w = r wv + r wm

(1)

Here I am the input source, and rw is the energy source. Step 3:Calculate the power source for a fiber optic system Ir w = rs0 , Pc− = −Ps−1 . . .

(2)

Here So is the source of the voltage, Pc is the storage value. Step 4: Calculate the bandwidth module. Ir w = Is0 , Pc = 0, Pc− = Ps−1 . . .

(3)

Step 5: calculate the output source in the fiber optic communication. A large number of Optical Network (ON) communication are guaranteed to address the issues have raised. Despite the organization’s enormous limitations, the Optical Network (ON) organization provides a typical foundation for the administration pattern. These communications are slowly getting fitted to deliver data transmission adaptively when and where needed. Optical fibers provide much higher

258

L. Bharathi et al. Start

Initialize Input Signal Source to fiber connection (Transmitter)

Optical communication through the response time in seconds

Calculate the output Bandwidth/Signal Select the Error

YES

And address to the

Vin == Vref

NO Vout>Vref

YES

Fiber to detector Connection (Receiver)

NO

Calculate the Error signal and address the source

Fig. 5 Flow chart of prescient control algorithms based optical network

transmission capacity than copper links and are less vulnerable to various electromagnetic barriers and other unfortunate effects. Therefore, it is a convenient vehicle for transmitting information on anything more than two or three megabytes per second over a distance of more than one kilometer. Similarly, there are favorable methods of accepting short distances (several meters to many meters), high velocity (every second or more gigabit) interconnections. Figure 5 gives the: Only a certain number of mods fit to spread through the waveguide. The transmission qualities of optical fiber links play an important role in determining a complete correspondence presentation.

4 Result and Discussion The organization application’s prerequisites have led to event twists and configured/ programmed optical organizations. In addition to the progress of programming characterization planning on the control plane and the included administration plane, naturally configured optical organizations need a brilliant calculation to advise the

Analysis of High Performance Optical Networks Using Dense …

259

control plane, for example, load adjusting, greenery systems administration, limit expansion. Table 1 Despite the drawn-out configuration, the transient arrangement is more solid on the design capability of optical organizations, as traffic requests are usually handled at short intervals between configuration phases and configuration phases. The certainty of the actual structure, the real level vulnerabilities when a sub-ideal asset or path can be given due to calculation or specific strategy Figure 6 gives: The target is to propose an accurate nonlinear impairment model first and then improve the network efficiency using the proposed model by allowing more requests to be established in the Network with various proposed optimization schemes. This proposal mainly focuses on configured transient optical organization, including steering calculations, frequency function, recovery recovery-based optical organization, traffic preparation process, and optical light path security. Table 2 the information that helps the light wave at that time, for example, optical fiber links in this structure, pass through the transmission medium. At present, it has reached the receiver stage, where the optical indicator demodulates the optical transporter and signals the electrical output at the electrical stage Table 1 Characteristics analysis of optical network Channels

Bandwidth kpbs

Bit-rate (Gbps)

Frame size (Per unit = 1)

Raising time (Time in Seconds)

CH1

53

24

98

1.75 s

CH2

22

10

18

3.50 s

CH3

24

12

22

3.89 s

CH4

32

10

76

1.15 s

CH5

43

10

4

1.94 s

Fig. 6 Efficient analysis of prescient control algorithm

260

L. Bharathi et al.

Table 2 Tabulation and calculation of Dense Wavelength-Division Multiplexing (DWDM) Fiber count (unit = 1)

Cable diameter (mm)

Cable net weight (kg/km)

Static rate

Modal bandwidth

2–6

9.01

98

109

159

8–12

9.06

115

108

199

115

119

399

14–24

10.1

Fig. 7 Wavelength analysis of Dense Wavelength-Division Multiplexing (DWDM)

Figure 7 gives the: Correspondence optical Network service for error rate performance using Dense Wavelength-Division Multiplexing (DWDM). The proposed Prescient Control Algorithm sensitivity performance error rate is low. Similarly, the existing methods Time Division Multiplexing (TDM) based Fuzzy Neural Network false rate performance result is low. The optical fiber correspondence system’s essential goal is to move an icon containing data (voice, information, and video) from source to object. Overall square chart of optical fiber correspondence system. The source gives the data to the transmitter as an electrical sign. The electrical phase of the transmitter drives the optical source to create an adjusted light wave transporter. Semiconductors are commonly used here as optical sources. The optical network optimization further into physical layer impairments aware network resource allocation address the routing, spectrum and modulation assignment for elastic optical networks Commonly used optical locators are photodiodes (Pin, tor rental slides), phototransistors, and photoconductors and so on. Similarly, the draw-out organization is represented as an arrangement of the assessment stage, zero to determine the ideal procedures for sending the foundation to arrange the assets’ allocation. Lacing and data transfer capability are the two most notable transmission features when optical fiber’s suitability for correspondence is broken. Different

Analysis of High Performance Optical Networks Using Dense …

261

vulnerabilities are systems such as direct disassembly, nonlinear deceptions, material assimilation and fiber twists.

5 Conclusion Optical Network (ON) supports a wide range of correspondence management, including private administration, administrative administration and multidisciplinary administration. Network communication of the mile to terminate the fiber optical correspondence network by accessing the center, metro and optical access organizations. Imperative fifth-age remote organization for new optical systems administration requirements such as high transfer speed, low latency, precise synchronization, and network cutting ability. The Dense Wavelength-Division Multiplexing (DWDM) prerequisite for high transfer speeds is many data through remote applications. However, the requirements for low inactivity and precise synchronization are essentially determined by applying the Fiber Optic Communication (FBC) and Composed Multi- Point (CMP). The prerequisite for network cutting is directed towards increasing asset usage for some random application. Each of these requirements tends to be in the supposed aligned optical wired Network. The survey is planned to show the emotional mechanical progress in fiber transmission and systems administration in getting speed and give viewpoints.

References 1. Yang, H., Yao, Q., Yu, A., Lee, Y., & Zhang, J. (2019). Resource assignment based on dynamic fuzzy clustering in elastic optical networks with multi-core fibers. IEEE Transactions on Communications, 67(5), 3457–3469. https://doi.org/10.1109/TCOMM.2019.2894711 2. Keykhosravi, K., Agrell, E., Secondini, M., & Karlsson, M. (2020). When to use optical amplification in noncoherent transmission: An information-theoretic approach. IEEE Transactions on Communications, 68(4), 2438–2445. https://doi.org/10.1109/TCOMM.2020.2968890 3. Saeed, N., Al-Naffouri, T. Y., & Alouini, M. (2019). Outlier detection and optical anchor placement for 3-D underwater optical wireless sensor network localization. IEEE Transactions on Communications, 67(1), 611–622. https://doi.org/10.1109/TCOMM.2018.2875083 4. Dehkordi, J. S., & Tralli, V. (2020). Interference analysis for optical wireless communications in network-on-chip (NoC) scenarios. IEEE Transactions on Communications, 68(3), 1662–1674. https://doi.org/10.1109/TCOMM.2019.2960339. 5. Celik, A., AlGhadhban, A., Shihada, B., & Alouini, M. (2019). Design and provision of traffic grooming for optical wireless data center networks. IEEE Transactions on Communications, 67(3), 2245–2259. https://doi.org/10.1109/TCOMM.2018.2885808 6. Guzmán, B. G., Dowhuszko, A. A., Jiménez, V. P. G., & Pérez-Neira, A. I. (2020). Resource allocation for cooperative transmission in optical wireless cellular networks with illumination requirements. IEEE Transactions on Communications, 68(10), 6440–6455. https://doi.org/10. 1109/TCOMM.2020.3010583 7. Zhong, J., Xie, W., Li, Y., Lei, J., Du, Q. (2020). Characterization of background-anomaly separability with generative adversarial network for hyperspectral anomaly detection. IEEE

262

8.

9.

10.

11.

12.

13.

14.

15.

16.

17.

18.

L. Bharathi et al. Transactions on Geoscience and Remote Sensing. https://doi.org/10.1109/TGRS.2020.301 3022. Li, C., Cong, R., Hou, J., Zhang, S., Qian, Y., & Kwong, S. (2019). Nested network with two- stream pyramid for salient object detection in optical network sensing images. IEEE Transactions on Geoscience and Remote Sensing, 57(11), 9156–9166. https://doi.org/10.1109/ TGRS.2019.2925070 Wu, S., Ren, J., Chen, Z., Yang, P., Li, H., & Liu, J. (2020). Evaluation of winter wheat yield simulation based on assimilating LAI retrieved from networked optical and SAR remotely sensed images into the WOFOST model. IEEE Transactions on Geoscience and Remote Sensing. https://doi.org/10.1109/TGRS.2020.3038205. Yang, Z., Guan, X., Luo, B., Yin, L., Gao, G., Chen, J. (2019). Ambient light suppressed free space optical communication system based on FADOF. In 2019 18th International Conference on Optical Communications and Networks (ICOCN), Huangshan, China, 2019 (pp. 1–3). https://doi.org/10.1109/ICOCN.2019.8934667. Li, Z., Wu, C., & Wang, Z. (2011). Polarization temporal signal processor for ultrahigh-speed optical communication applications. In 2011 Optical Fiber Communication Conference and Exposition and the National Fiber Optic Engineers Conference, Los Angeles, CA, USA, 2011 (pp. 1–3). Huang, Y., Wang, X., Wang, K., Zhang, D., & Dai, B. (2017). A novel optical encoding scheme based on spectral phase encoding for secure optical communication. In 2017 16th International Conference on Optical Communications and Networks (ICOCN), Wuzhen (pp. 1–3). https:// doi.org/10.1109/ICOCN.2017.8121476. Yunzhu, M., Zhengbin, L., & Anshi, X. (2006). Novel optical filters configuration for dispersion compensation in high-speed WDM optical communication systems. In 2006 First International Conference on Communications and Networking in China, Beijing, China (pp. 1–5). https:// doi.org/10.1109/CHINACOM.2006.344819. Rahaim, M., & Little, T. D. C. (2017). Interference in IM/DD optical wireless communication networks. IEEE/OSA Journal of Optical Communications and Networking, 9(9), D51–D63. https://doi.org/10.1364/JOCN.9.000D51 Kavitha, G., & Sundararajan, J. (2012). An intelligent routing protocol for MANETs using radial basis function neural networks. European Journal of Scientific Research (EJSR), 89(3), 342–349. ISSN 1450-216X. (Impact Factor 0.736). Bharathi, L., Sangeethapriya, N. (2019). Attribute table based energy-efficient and qos-of multipath routing protocol using in loss- free optical burst switching networks. Journal of Advanced Research in Dynamical and Control Systems, 11 Special Issue. (ISSN 1943-023X). Bharathi, L., & Sangeethapriya, N. (2020). An improved link stability based on swift exploring packet ratio using expected time matrix in wireless sensor network. Journal of Computational and Theoretical Nanoscience, 17(12), 5528–5534. SangeethaPriya, N., & Bharathi, L. (2020). Optimal multilevel link extend umpiring routing based secure routing protocol for improving privacy standard in wireless sensor networks. Journal of Computational and Theoretical Nanoscience, 17(12), 5543–5549.

Wireless Sensor Network to Improve Security Performance and Packet Delivery Ratio Using FCL-Boost Based Classification Method N. Sangeethapriya, L. Bharathi, S. Jagan Mohan Rao, and A. N. L. Harisha

1 Introduction This study began with a useful question on improving wireless sensor networks’ network life cycle using reinforcement learning enhanced power with sensor terminals. The FCL-Boost algorithms to improve energy efficiency, problem-solving has been the main area of study in the various literature that has been discussed over the past few decades. Therefore, it is expected that the sensor nodes will improve and enable sleep and wake-up mechanisms to exploit their energy to improve the networks’ life cycle. In most cases, due to the multi-hop communication, it is a heavy load to the nearest base station as the senior center because they are centered in the middle of the base station and stay away and arrange the information sent to the remote sensor base station. The problem in this scenario is the problem area where the Sensor Node (SN) sends its information, such as different hub information near the center of the event tank. Of these contributions to the total pollution reduction worn by the remote sensor at the exhibition. Wireless Sensor Network Development and Problem Areas, Generally speaking, the head of the universal tank is required for each beam, information collection, and information transmission can’t afford to take a long time to cross a long lifespan and uses little middle. Therefore, suggest learning support and data weight calculations. A WSN sensor is a self-organized network of nodes that show the Fig. 1 communicate with a group of nodes that send a special node activator into their environment, communicate approximately spontaneously through the unconnected wire, fixed or N. Sangeethapriya · L. Bharathi (B) · S. Jagan Mohan Rao · A. N. L. Harisha Department of ECE, Ramachandra College of Engineering, Eluru, Andra Pradesh, India e-mail: [email protected]

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 V. K. Gunjan et al. (eds.), Modern Approaches in Machine Learning and Cognitive Science: A Walkthrough, Studies in Computational Intelligence 1117, https://doi.org/10.1007/978-3-031-43009-1_22

263

264

N. Sangeethapriya et al.

Fig. 1 Architecture of WSN communication user to server [1]

scattered over a specific geographic area, and a set point of data, where the sensor node company is present in a field. The transfer of collected data done from time to time or In light of the application running’s occasion-based nature. One end of the container, WSN, and end-client organize (e.g., a neighborhood or the Internet) is an interface between the extension between and is more than one system. The client to determine the sort of information gathered, for instance, through a water tank in the system, can be mentioned to send to another hub. For illustrations, a wireless sensor network type’s one basic program engineering. The sensor center is located at the end of the plot area, and area Duty regarding acquiring the information gathered through the sensor field. Regularly, the tank in the data venture that has been gathered and composed in such an Away, only relevant data is sent to the customer. It can receive orders and run from its internal system (customer). The information collected is processed and broken down by the customer. A wireless sensor is a small electronic device that can measure body level (such as temperature, light, pressure, etc.) and transmit it directly or through other sensor nodes to a set center that acts as a router. Given the advancements made in microelectronics, wireless transfer technology and software make smaller scale sensors that can create a few cubic millimeters. It can work the module in the system at a sensible cost. The sensor hub incorporates four fundamental units of two altered hand-held acquiring implies. Sensor and a Simple Computerized Converter (SCC): These have two subunits. The sensor transforms digital measurements of environmental factors, and Analog Transfer Collects the information. It consists of two units: the data storage unit that allows other sensors to be used for collection tasks, and the processor is responsible for processing data and control procedures. The unit corresponds, its

Wireless Sensor Network to Improve Security Performance and Packet …

265

capacity is data transfer and compilation. It has a transmitter/receiver pair. It is the internal communication system that, in this case, allows this problem RW (Radio waves). In any case, solar panel power units’ continuous testing attempts to answer extended sensor life. Mobilization. Alternatively, this task can be used to move the process. Location detection system to complete the hub. Alternatively, this application or possibly provides the required terrain data through guidance. This chapter is discussed above the wireless sensor data security and processing explains this topic is covered networks.

1.1 Machine Learning Algorithm What can be applied to calculate under the jurisdiction of Artificial Intelligence (AI), in the past, based on the guidelines on the use of naming, it is expected that in the future new information is realized. Known starting from the investigation of the prepared data set and to learn to calculate, create a derived ability to make predictions about the output self-esteem. The framework can give the focus to any new contributions after adequate preparation. Calculation of learning can also adjust the appropriate model, compare the rate of correct the model. Interestingly, if they want to use it to prepare the data, calculating one of the groupings and the solo AI name is not available. The concentrated visual framework will learn how it is possible to induce the ability to portray the hidden structure of the non-labeled information. A framework does not make sense of the accuracy rate’s return; it examines the information and pulls out the non-labeling of information from a data set that can attract derived from the hidden structure.

1.2 Security Level of Methods By radio communication of the mobile remote system, the simple screen can attack the attacker without being associated with the actual system. Attack, will pass through the access systems in the range of remote systems that are not protected. Some experts recognize any channel and the remote system interface card design security coordination degree.

1.3 Wireless Sensor Network Behavior The goal is to change the universal arrival operation setting classification to calculate the remote sensor system’s insight to show a sharp action. The quality of the remote sensor system provides several difficulties. For example, the number of sensor hubs, thick tissue to change the geographical structure is enormous, including power,

266

N. Sangeethapriya et al.

computing, storage, and the corresponding ability is a specific constraint asset. Rules of all applications, vigor effective needed to run on a remote sensor system, universal but is violent, “tuned” to change conditions and settings. Without the application extension, action-filled with disposed insight in the middle between the other people and performance.

2 Related Works The Personal Computer systems and the widespread use of online electronic information and safety appeals have called for a solid framework for [1] cross-discovery. Muss rappels traitor de aboard Human issues, Society General de la combination affiliate unit, etc. proposes taxonomies [2] methods. Elfin, rational Les attends equal value, etc., ready to reduce particular la [3, 4] difficult single generalizer fewer results. This technology’s principle idea is to use the Relative Entropy (RE) principle to evaluate organized traffic to be distorted and identify the system’s characteristics. Depending on the model, is also planned. Relative entropy is to expand increment the location rate, counterfeit the [5] alarm rate, which is more accurate and can reduce the inherent logical discrepancy between the most estimated errors in practice. Lincoln Laboratories Assessment Information Index shows that identification results can come at high levels and speeds due to the low false alarm rate strategy. Assault location innovation [6] depends on latent observing of traffic to implement a detective system based on hidden, active measurement, real-time detection without intrusion detection, and as much classification as possible for attacks. The contribution is calculated using the probability of responding to a request measured from the delay time series that relies on the original entropy function/echo. However, this method [7] evaluation showed a substantial number of false positives. Then used to substantially reduce the probability of the number of false alarms using Hausdorff’s distance enrichment taking place in the time series. Using data processing [8] technologies such as nerve blur and radial base Support Vector Machines (SVMs) for IDs to help obtain a new method of high detection rate. The proposed technique [9] has four main steps: First, the K-means set is used to create different training subsets. Then, according to the training subcommittee obtained, different nerve ambiguity models were trained. Following that, a vector of SVM classification is generated, and finally, the radius is made to determine whether SVM classification infiltration has occurred or not. To illustrate this new method’s compatibility and efficiency, the results of the experimental file dataset were demonstrated. Experimental [10] results show that a new method is a really good blood pressure neurological network, multi-class SVM, and other well-known methods, such as the Columbia model based on the decision trees’ sensitivity and accuracy, especially the result. Hypothesis [11, 12] from straightforward programming duplication provides a fascinating understanding of the question and enables us to determine O | E | Registration | E | the closest answer to time and self-esteem is to double the K-focal best

Wireless Sensor Network to Improve Security Performance and Packet …

267

value. The regulation in each area of information is implemented due [13, 21] to the Dumpster rule of combinations of these possible components. As a result, the first two datasets’ order showed surprisingly high accuracy while other well-known strategies were not considered rigorous. The third set of data groups is non-digital and difficult. Still, completed by giving the structure and quality capabilities a great result to get the [14] combination for a carefully structured and accurate characterization. In all cases, the DS sources strategy gives almost the same performance as other well-known calculations. Still, the quality of production accuracy gives cost processing and expansion of new features. In general, the [15, 22] results suggest that the gives a large-scale capacity structure and size to mechanize the compatibility of calculating the structure of a reasonable [16] construction system and complex order issues. In this way, the success of distinguishing interference under these conditions, unfortunately, requires a significant degree of accuracy and productivity with a preempted goal. Many interfering [17, 23] detection patterns in writing do not recommend continuous responses to manage the barriers mentioned above. This experiment and Denial of Service DoS encourages us to propose a lightweight interference recognition framework for attack locations. Even an isolated [18] occasional source (for example, transportation arrangement) thinks about it alone when facing big data challenges. Attempts to use more diverse information sources represent a challenge that deserves a lot of attention with big data. Innovative infiltration detection of bulk [19] data can lead to the realization of these large diversified data challenges. This article investigates the issue of diversified information, especially to the degree of big difference. This report is a precursor to an infiltration detection system. Details of possible attacks and [20–23] reactions were detected in the water at the beginning of a conversation about various components. In the long run, it will hint at the new concept of IPS (Intrusion Prevention System) and possible dialog ways to improve the existing structure. In this option, up to two unique techniques, particularly FCLBoost.L1 and be FCLBoost.L2, F-score calculation for judgment in the light of implementation, highlighting grown and ready to collect the classification can be upgraded. 1. It is done. (This is a makeover whose process has at least three distinct classifications; finally, the classification of the three distinct classifications combine theater company. Along these lines, it is discharged from the production process to ensure that information can be meaningless. Equipment classification Yes, the project took a little later redesigned to be considered. Are a mixture of three types of unique types of collections in the long (The use rate of e-orphan classification is reduced by cost. The only applicable feature is the use of F-score recovery, including judgment calculations.

268

N. Sangeethapriya et al.

3 Materials and Methods The proposed system to overcome the above problems efficiently. Recent detection minimizes show the Fig. 2 first discriminatory algorithm (FCL-Boost) algorithm using detection. It is adding data to a user request that the L1 level does not have a one-time data response. The second level of information to users on a server in the data travels safely. The stream performed the activity in Fig. 2. Right off the bat, the records to be utilized in the examination were gathered. The highlights were chosen with the F-Class include choice calculation. At last, the information is arranged with various classifiers, and their exhibitions are determined. When these assignments are performed, the hardware has delegated a classification and presentation that is resolved at all levels and positions.

3.1 Collection of Data From the Classifier Outfit by engaging each classifier is a classifier to provide a safer, more stable frame made meter the classification framework is N ‘may be single or double. Although characterized as indicated by the vector component, each classification yield creates self-esteem for each vector element 1. (Delivery yield respected figured out, the group classification yield resolved by votes of Opportunity closed on the classification of the even number is selected, the normal classification of the estimated adjustment, and the choice to solve the troupe classification). (A process it applies to all components of the carrier) Expanded the range close to zero, the system image and nearest neighbor classifier start. (Self-esteem when you’re far away from 0, multiple vectors, each with different information, the classified into consideration gives birth order). The extension range is limited to 0.01 in advance. After examination, the nature of the outskirt and usage of the best execution framework has been resolved. Supported vector machines have better AI calculations. The Support vector machines try to isolate a straight line and a dotted line from a set of data. The basic error information in the (Support Vector Machine) SVM calculation and an option must be recognized. Solution analysis is used in the Gaussian or Extended Conditional Work Random Forest (RF) area (RF). (The power box is set to 100 targets somewhere in the control box 1 is limited to the range and can be expected to be completed).

Wireless Sensor Network to Improve Security Performance and Packet …

WIRELESS SENSOR NETWORKS

USER REQUEST

DATA RESPONSE

269

DATA COMMUNICATION

SECURE DATA TRAVEL WEIGHT (Wgt)

ROUTE ENERGY

FC-BOOST SUPPORT FCL-Boost First Level DATA RECEIVED

FCL-Boost Second Level

Fig. 2 The proposed system’s architecture to the data security and Route energymethod using the FCL-boost algorithm

3.2 Classifier-Based Ensemble Classifier: FCLBoost.L1 Method Simulation Steps Fig. 3. Thus classified in the classifier (KNN) is shown in detail in this first set of data (a). The second step is to repeat the first edition featuring and selected (KNN is categorized in a similar classification). In a third step, perform the first and second component determination repeating the classification (KNN) in the same category. Three different activities were divided in the US, but the same classification (KNN) was found. These three outcomes are consolidated to frame a kclosest neighbor gathering. Rehash a similar procedure Random ForestRF and SVM. Finally, Q Neighbor Set, RF [Random Forest] Group, and SVM Group combine into a single, integrated classification.

3.3 Functional Integrated Classifier: FCLBoost.L2 The steps of this method are shown in Fig. 3. Therefore, the data set (a) is first classified according to each classifier (adjacent ones, RF, and support vector machine k). These three categories are integrated type 1. The second aspect is integrated to get, the first aspect is selected, and the Handling is done in the initial step. In a third step, the first and second characteristic choice advance is executed together, at that point rehashes the primary formula. Order for bunch 1, 2, and 3 are combined to form a comprehensive recognition.

270

N. Sangeethapriya et al.

3.4 Route Energy Path Collaboration identifies the overall vitality of the connected hub and has the best vitality process of photographing a priori information on the hub’s surface position. The S- surface position hub will be occupied in the corresponding position to the closing of the opportunity. It updates the underwater neighbors to adopt the center of the then to the surface position of the replacement to avoid loss of information. From that point on, the hub’s water will push the particular data to the storage hub and take another surface position of the hub. Finally, in advance of the hub to the surface position data, information is most transmitted in proficiency mode vitality most minimal jump contemplated the least limit of the separation process. In the end, information is loaded to pick up it is sent the shell from the hub. Besides, a once-over having information log is kept up by load center point, containing data about which data gets sent from which explicit center point and at what repeat of time term the information was gotten, going before taking care of the proportionate the shell. The particular shell that goes about as pad stores the data advanced through the stack center point. Additionally, they are using divided data strong trade shows that achieve a reliable imperativeness amazing transmission of data in submerged sensor sorts out. Figure 4 expressions a flowchart of the Foremost Classification Level FCLBOOSTmethod algorithm and stages. Fig. 3 Data flow of FCL-boost ensample algorithm

Wireless Sensor Network to Improve Security Performance and Packet …

271

Algorithm: Step 1: Assume that all the variables by the device are a prediction. Step 2: Predict each tracking error from the average (latest forecast). Step 3: To find the completely separable error, find the extractable value variable. This latest forecast is considered. Step 4: Predict the classification (recent prediction) error from each observation on the marking page. Step 5: Repeat steps 3 and 4 to enlarge/reduce the scope function. Step 6: Take the weighted average of all the categories to come up with the final sample 3.1 Predicting is the client input distance based on the information obtained from the request example and. 3.2 The clients requested location and path code distance will be included in the ordered information Input: source Client FileCF(text), LP (Location Path), Transfer dataT, Secure server SS. Output: securedata transferSDT. Start The source inclient to load the fileinfoFCL-Boost-data, Read Location Path LP, the nearest secure server verifies and clientNode connects level. Make NC-Req path CF -Req = {Starting IP, Ending IP} data verify to location and PC-Boost active Map PathCF-Req IF-Node PATH VERIFY runs Node- PATH CF –Req Node match active to PC-Boost Else SERVER NC -Req to its select another path. Go to for current location and server map location active. Find the Alder route on the detection of security. Stop

272

N. Sangeethapriya et al.

4 Result and Discussion The motivation behind this work is to build up another calculation to improve the general acknowledgment execution. The quicker, perform better execution, even though the calculation (FCL-support) presentation is comparable to the irregular interior framework with less exertion. Because of this strategy, there are two adaptations of the F-score include choice calculation (FCLBoost.L1 and FCLBoost.L2). (FCLBoost.L1 is broad recognition and development of a single taxonomy, Level 2) and FCLBoost.V2 into at least three types (Level 1). (Developed according to the FCL-Boost algorithm, double-select the features of the dataset using the F-level feature selection algorithm.

4.1 Implementation and Experimental Results In this section, implement the main goal proposed system through the establishment and use of a parody of notification of the application is Network security.NS2Network simulator. The name of the Foremost Classification Level algorithm method (FCLBoost) is. According to this method, the selection and classification of different classifiers or F level is the same assortment. The next step is to create a unified classification. Two versions, FCLBoost.L1 (Level 1) and FCLBoost.L2 (Level 2), under the same or different classifications have been developed.As per the outcomes got, the outcomes are predictable with writing reports. FCLBoost.L1 and FCLBoost.L2 for exactness rate are the primary higher than FCL speeding up the calculation. FCLBoost.L1, 1-SVM joining strategy is the ideal way a higher position than the other upgrade calculations. The time has come to guarantee that it is utilized to re-check the aftereffects of various informational collections were gotten. (To build the FCL boosting calculation as the dispersion of the information collection is summed up in the consequences of three distinct informational collections that appeared in Table 1 is gotten from the reanalyzed. (Examination for correlation. According to the normal execution of the calculation, these outcomes are the best beneficial thing is FCLBoost.L1 level two bound together calculation.

4.2 FCL-Boost Network Performance Support vector shows the Fig. 5 machine for measuring network performance of data protection system segment SVM and RF [Random Forest] and key nearest neighbor [KNN] and finally Foremost Classification Level [FCL-Boost] algorithm, method. In X- axis is the method of comparison with the unit of measurement algorithm

Wireless Sensor Network to Improve Security Performance and Packet …

273

Table 1 Parameter values Parameters

Value

Simulation

Cygwin

Propagation model

Detrimental in radio communication systems

Area

800 m × 800 m

Broadcast area

50–250 m

Transfer pattern

UDP, CBR

Mobility model

Random mobility

Transfer per packet

512 bytes

No of nodes

40

and y-axis. The FCL-Boost network performance method differs from the focus of different data. Better than KNN. Tertiary performance ratio FCL-Boost ensample. Rate network performance increase for FCL-Boost 85% compared to SVM, RF, and KNN method.

4.3 Time Complexity When one model is used, it is usually taken as a system identifier by the confirmation group collected from the information, while the other is used to produce the package. In these ways, I can collect different samples of my test information coding. The choice of technology, for example is due to accidental network security confirmation team-related training materials. My best example is drawn by a Fig. 6 in the information list. Show the Fig. 6 Evaluation SVM [Supported Vector Machine] and RF [Random Forest] and KNN [FCL-Compensation] Algorithm, Time Performance Team Performance Y Performance Algorithms in State X Center Comparison Times. Methods Classification is proposed to measure the life extension at critical moments and critical moments when it is recommended that the radio product SVM 70 s RF60 s products one or two moments less incredible. KNN 55 s, this structure than the previous strategy. Proposed FCL Accelerated Calculation, 50 s Study Program Framework, Detection Best Performance FCL-Boost Half on Data Security Broadcast System, KNN 55%, and Random Forest 60%. FCL-boost performance 50% better timing for accessing the main goal of the proposed system.

274

N. Sangeethapriya et al.

Fig. 4 Flow chart of the FCL-BOOST algorithm process

4.4 Average Throughput The number of typical open doors, give each time depends on the particular purpose. It how fast ratio characterizes the wheel hub, it can send the information through the system. Production normal throughput rate of the messages is sent via the wireless sensor network’s corresponding multiplexer for a massive time. Table 2, the average output rate test, demonstrates the recommended configuration in the network. It would be the most proposed comparison of the number of existing node values in node processing. The demo check is shown in Fig. 7. In this profile, the active test X is indicative of the Number of nodes and y-axis to indicate throughput analysis in the percentage

Wireless Sensor Network to Improve Security Performance and Packet …

275

Fig. 5 FCL-boost Network performance

Table 2 Throughput analysis Number of nodes

SVM %

KNN %

RF %

FCL %

10

28

31

34

45

20

39

48

50

55

30

45

51

55

65

40

55

64

70

75

of its turning speed count. These are dark orange and RF blue, and SVM from KNN. FCL compares best with yellow light and shade, for which the frame will make the current frame elite study recommended level.

4.5 Packet Transfer Ratio A determined fixed number of pocket loop regions of the beam sent from the source end depends on being able to be the target rear. The estimated speed of the pocket movement, Pocket Transfer Rates are dark orange and RF blue and SVM from KNN. FCL compares best with yellow light and shade, for which the frame will make the current frame elite study recommended level. Figure 8 shows the packet transfer rate. Like the bundle number of x and nodes, this number is the number of talking centers centered on the y-axis pocket transfer rate. The color blue (SVM) and color orange (KNN), and emulsion (radio frequency)

276

N. Sangeethapriya et al.

Fig. 6 Time complexity

color oversee the percentage of yellow color compared to the system with the highest transfer rate proposed system from the problem in these previous methods. Multiple Lead Classification Level to support Foremost Classification Level has the best method (FCL).

Fig. 7 Throughput for compares of method

Wireless Sensor Network to Improve Security Performance and Packet …

277

Fig. 8 Packet transfer ratio

5 Conclusion The Foremost Classification Level [FCL Boost] might be created that the best computing. The strategy is not much progress has been just that, this gives quickly, high- accuracy, high-exactness, and quick outcomes. An especially invaluable situating calculation is valuable in a variety of system security data. In this way, FCLacceleration might like it. FCL-Boost has lower, and steps accuracy ratio comparison is better and contains more than various computational calculations. Thinking about these attractions, FCL-acceleration may soon be the most frequently used calculation. FCL stimulus calculation is ideal for use as a symbol of clinical research in living, in-depth learning, and used consistently in computer communication. At least three classifications of FCL stimulus can be used. Also, FCL-Boost.L1 has an incentive level that can be used as an isolated classifier for FCL. It is FCL-Boost. The L1-specific bit reaches the highest standalone classifiers. The F score shows the judgment calculation to make this favorable position. By adding different highlights, similar details can be deciphered in unexpected ways. Once the opportunity closed, the classification performed in solid, FCL accelerated increments. Likewise, it is recommended to use powerful classifiers for calculations. Equipment classification often participates in the weak classification to draw a solid classification. The strong of FCL-Boost. This will say that FCL help is another approach to make a troupe arrangement. Improvement of the rate net presentation for FCL-Boost 85% compared to SVM, RF, and KNN method. Best Performance FCL-Boost Half on Data Security Broadcast System, KNN 55%, and Random Forest 60%. FCL-boost performance 50% better timing for accessing the main goal of the proposed system. A superior outfit classifier can be made with an incredible classifier, and the F-classifier includes choice calculation.

278

N. Sangeethapriya et al.

References 1. Borji, A. (2007). Combining heterogeneous classifiers for network intrusion detection. In Advances in computer science - ASIAN 2007. Computer and network security (pp. 254–260). 2. Zouari, H., Heutte, L., Lecourtier, Y., & ALimi, A. (2002). “Un panorama des méithodes de combinaison de classifieurs en reconnaissance de formes”, (French) [An overview of classifier combination methods in pattern recognition]. In Proceedings of the RFIA’2002 (Vol. 2, pp. 499– 508). 3. Zhang, Y., Han, Z., & Ren, J. (2009). A network anomaly detection method based on relative entropy theory. In Second international symposium on electronic commerce and security (pp. 231–235). 4. Hamamoto, A. H., Carvalho, A., Sampaio, L. D. H., Abrao, L., & Mario, T. L. P. (2018). Network anomaly detection system using genetic algorithm and fuzzy logic. Expert Systems with Applications, 92, 390–402. 5. Paz, J. S., & Roman, D. T. (2015). On entropy in network traffic anomaly detection. In 2nd international electronic conference on entropy and its applications. 6. Maklin, C. KL divergence python example. Towards data science. https://towardsdatascience. com 7. Labit, Y., & Mazel, J. (2008). HIDDeN: Hausdorff distance based intrusion detection approach DEdicated to networks. In The third international conference on internet monitoring and protection (pp. 11–16). 8. Dua, D., & Graff, C. UCI machine learning repository. In Irvine. CA: the University of California, School of Information. 9. Wang, G., Hao, J., Ma, J., & Huang, L. (2010). A new approach to intrusion detection using Artificial Neural Networks and fuzzy clustering. Expert Systems with Applications, 37, 6225– 6232. 10. Lima, C. F. L., Assis, F. M., & de Souza, C. P. (2012). A comparative study of the use of Shannon Réinyi and Tsallis entropy for attribute selecting in network intrusion detection. In Intelligent data engineering and automated learning (pp. 492–501). Berlin: Springer. 11. Hochbaum, D. S., & Shmoys, D. B. (1985). A best possible heuristic for the k-center problem. Mathematics of Operations Research, 10(2), 180–184. 12. Tavallaee, M., Bagheri, E., Lu, W., & Ghorbani, A. (2009). A detailed analysis of the KDD CUP 99 data set. In Second IEEE symposium on computational intelligence for security and defense applications. 13. Chen, Q., Whitbrook, A., Aickelin, U., & Roadknight, C. (2014). Data classification using the Dempster-Shafer method. Journal of Experimental Theoretical Artificial Intelligence. 14. Surathong, S., Auephanwiriyakul, S., & Umpon, N. T. (2018). Decision fusion using fuzzy Dempster-Shafer theory. In Recent advances in information and communication technology (pp. 115–125). 15. Xu, L., Krzyzak, A., & Suen, C. Y. (1992). Methods of combining multiple classifiers and their applications to handwriting recognition. IEEE Transactions on Systems Man Cybernetics, 22(3), 418–435. 16. Tchakoucht, T. A., & Ezziyyani, M. (2018). Building a fast intrusion detection system for highspeed-networks: Probe and DoS attack detection. Procedia Comput Sci., 127, 521–530. 17. Zuech, R., Khoshgoftaar, T. M., & Wald, R. (2015). Intrusion detection and big heterogeneous data: A survey. J Big Data., 2, 3. 18. Sahasrabuddhe, A., et al. (2017). Survey on intrusion detection system using data mining techniques. Int Res J Eng Technol., 4(5), 1780–1784. 19. Dali, L., et al. (2015). A survey of intrusion detection systems. In 2nd world symposium on web applications and networking (WSWAN) (pp. 1–6). Piscataway: IEEE. 20. Scarfone, K., & Mell, P. (2007). Guide to intrusion detection and prevention systems (idps). NIST Special Publication, 2007(800), 94. 21. Debar, H.: (2000). An introduction to intrusion-detection systems. In Proceedings of Connect, 2000.

Wireless Sensor Network to Improve Security Performance and Packet …

279

22. Bharathi, L., & Sangeethapriya, N. (2019). Attribute table based energy- efficient and qosof multipath routing protocol using in loss- free optical burst switching networks. Journal of Advanced Research in Dynamical and Control Systems, 11. ISSN 1943–023X 23. Bharathi, L., & Sangeethapriya, N. (2020). An improved link stability based on swift exploring packet ratio using expected time matrix in wireless sensor network. Journal of Computational and Theoretical Nanoscience 17(12), 5528–5534. 24. SangeethaPriya, N., & Bharathi, L. (2020). Optimal multilevel link extend umpiring routing based secure routing protocol for improving privacy standard in wireless sensor networks. Journal of Computational and Theoretical Nanoscience, 17(12), 5543–5549.

To Analyse the Impact of Integration of Wind and Solar Power Generation System for Uttarakhand, Haryana and Rajasthan: A Scope of Machine Learning Himanshu Giroh, Vipin Kumar, and Gurdiyal Singh

1 Introduction Energy sources that naturally replenish themselves and do not run out over time are referred to as renewable energy sources. Solar, wind, hydro, geothermal, and biomass energy are some of these resources. Solar energy is produced by photovoltaic panels or concentrated solar power plants, which harness solar energy and turn it into electricity. By employing wind turbines to capture the wind’s energy, we can produce wind energy [1]. When the wind blows, the turbine’s blades rotate, driving a generator to generate power. Hydroelectric power is created by turning a turbine with the force of flowing water, which then drives a generator to provide electricity. Dams, which hold water in reservoirs and release it through turbines, or smaller, decentralised hydroelectric systems, can be used to do this. Geothermal energy, which can be utilised to produce electricity or to heat and cool buildings, is created by capturing the heat from the Earth’s core. Burning organic resources, such as wood or agricultural waste, to create heat or power is known as biomass energy. Reducing our dependency on fossil fuels, which play a significant role in climate change and air pollution, is one advantage that renewable energy sources provide. A sustainable choice for supplying the world’s expanding energy needs, renewable energy sources H. Giroh (B) · V. Kumar · G. Singh Department of Electrical Engineering, UIET, Maharshi Dayanand University, Rohtak, Haryana, India e-mail: [email protected] V. Kumar e-mail: [email protected] G. Singh e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 V. K. Gunjan et al. (eds.), Modern Approaches in Machine Learning and Cognitive Science: A Walkthrough, Studies in Computational Intelligence 1117, https://doi.org/10.1007/978-3-031-43009-1_23

281

282

H. Giroh et al.

also have the potential to offer a clean and limitless source of power. Fossil fuels, which are a major cause of climate change and air pollution, have received a lot of attention recently as viable alternatives to renewable energy. Renewable energy sources, including solar, wind, hydropower, and geothermal energy, have the potential to provide a clean and limitless source of energy, making them a viable choice for supplying the world’s expanding energy needs [2]. The switch to renewable energy does not, however, come without difficulties. The expensive upfront costs of putting these technologies into practise are one of the key barriers to the widespread use of renewable energy. For instance, installing solar or wind energy systems demands a substantial financial commitment, which may be prohibitive for private citizens, companies, or governments. Additionally, building new infrastructure for renewable energy sources, such as transmission lines and storage facilities, can be expensive and time-consuming. In recent years, there has been a growing focus on increasing the use of renewable energy in order to reduce reliance on non-renewable energy sources and mitigate the negative impacts of climate change. Many countries and businesses have set ambitious goals to transition to renewable energy sources as part of their efforts to reduce greenhouse gas emissions and become more sustainable.

1.1 Energy Crisis and Renewable Energy Use The term “energy crisis,” frequently referred to as “the global energy dilemma,” is used to characterise the difficulties in addressing the world’s rising energy needs in a way that is sustainable and reasonably priced. Population expansion, economic progress, and an increase in the usage of energy-intensive technology are some of the causes that are causing the complicated problem known as the energy crisis. The usage of fossil fuels, such as coal, natural gas, and oil, which are a significant source of greenhouse gas emissions and are a contributing factor to climate change, is one of the key causes of the current energy crisis. Price volatility and supply interruptions have resulted from the rising energy demand and the constrained supply of fossil fuels, which can have a detrimental effect on the economy and society. Renewable energy sources, such as solar, wind, water, geothermal, and biomass, are naturally replenishing and do not deplete, unlike non-renewable energy sources such as fossil fuels. The increasing use of renewable energy has become a global priority due to concerns about climate change and air pollution, as well as the finite nature of non-renewable energy sources. In addition to individual efforts to transition to renewable energy, there has also been a push at the policy level to increase the use of renewable energy. Many countries and businesses have set ambitious goals to increase their use of renewable energy as part of their efforts to reduce greenhouse gas emissions and become more sustainable. Governments have also implemented policies and incentives, such as tax credits and subsidies, to encourage the adoption of renewable energy. Overall, renewable energy sources offer a sustainable and costeffective alternative to non-renewable energy sources, with numerous environmental and economic benefits. While there are challenges to increasing the use of renewable

To Analyse the Impact of Integration of Wind and Solar Power …

283

energy, the transition is becoming increasingly viable as technology improves and costs fall. Policy efforts at the national and international level are also helping to accelerate the transition to renewable energy [3].

1.2 Integration of Machine Learning Machine Learning can analyse the impact of integrating wind and solar power generation systems in Uttarakhand, Haryana, and Rajasthan by collecting and processing data, assessing resource potential, forecasting power generation, evaluating grid integration and stability, optimizing energy management, assessing environmental impact, and providing decision support. By leveraging Machine Learning ‘s capabilities in data analysis, and simulation, policymakers, energy planners, and stakeholders can gain valuable insights to inform decisions, maximize renewable energy utilization, minimize costs, and promote sustainable development in these regions [27].

2 Renewable Energy Sources Distribution Conventional energy is the main source of electricity generation but it produces a large amount of carbon and harmful gases in the environment. These gases are the main concern now a day in the world which damages the greenhouse system of the environment and increase Global Warming. Renewable energy sources have great potential to fulfil our energy demand without pollution. This is only solution of Global Warming and pollution. Power generation from renewable sources is on the rise in India, with the share of renewable energy in the country’s total installed capacity rising from 7.8% in 2008 to around 13% in 2014 (IREDA 2014). India now has about 36.4 GW of installed renewable energy capacity. Of these, wind is the largest contributor and stands at around 23.7 GW of installed capacity making India the world’s fifth largest wind energy producer. Small hydro power (4.1 GW), bio-energy (4.4 GW) and solar energy (4 GW) constitute the remaining capacity (MNRE 2015). It has been reported that in terms of electricity generation, approximately 70 billion units per year is being generated from renewable sources (MNRE 2014). Figure 1 below shows the renewable energy mix in the total installed capacity in India. This diagram clearly states that 79% energy comes from coal and gas, only 5.6% energy generate from renewable energy source. So, there is vast scope of renewable energy in India. Solar and wind both systems depend on environment condition so that we cannot depend on only one system. For hybrid of energy system, there are many types of schemes such as series hybrid system, parallel hybrid system, switched hybrid system. They used MPPT techniques to maximize the power point from solar module. P&O algorithm is used to track maximum point. In this method voltage is periodically compare the perturbation with output power perturbation cycle. But this

284

H. Giroh et al.

Fig. 1 Share of renewables in total grid installed capacity

method varies with varying atmosphere condition. Incremental Conductance method is used to overcome this drawback. The battery used for hybrid systems are lead acid battery, Nickel Cadmium battery. Others-Super conducting magnetic energy storage is used but it is expensive. Under trends heading, Hybrid renewable energy have great potential of research and development. Solar, wind, and hydroelectric power are examples of renewable energy sources that have grown in importance recently as a way to lessen our dependency on fossil fuels and lessen the damaging effects of carbon emissions on the environment. In comparison to conventional fossil fuels, these sources are more plentiful, cheaper, and emissions-free. The broad use of renewable energy sources still faces obstacles, including logistical and technological ones as well as the requirement for cutting-edge storage options to deal with some of these sources’ sporadic nature. Despite these difficulties, it is certain that renewable energy will play a significant role in our future energy mix, and work is being done to create more efficient and affordable solutions [4].

3 Recent Method and Techniques Used in Hybrid Renewable Energy System Amer et al. [5], this paper focus on optimization of power of Hybrid Renewable Energy System. For minimize the Levelized Cost of Energy (LEC), Particle Swarm Optimization Technique (PSO) is used. The proposed algorithm minimizes the cost fitness function of all sources of Renewable Energy. They achieve mainly two goals

To Analyse the Impact of Integration of Wind and Solar Power …

285

(1) Cost Reduction (2) maximize the Power Generation. They proposed a modified PSO algorithm and optimize problem, with the constructed objective function to be used in choosing how many powers will be generated in HRES in this research. PSO appears in the core of it, searching to find the optimum power generated. This routine is programmed under MATLAB software. Apply this algorithm they minimize LEC 0.0030277 e/kW and the power generated from the wind turbine is 1.0185 kW, while the power generated from the PV module is 0.23153 kW [5]. Andoulssi et al. [6], this paper modelling the PV generator, DC/DC buck converter and DC motor pump for HRES. PV generator used array of PV cell and implement IV equation of diode. Similarly buck converter and DC motor model is proposed and simulate on 20SIM. The simulation results indicated also a good performance of the controller. The overall system stability was studied and showed that the system is not of non-minimum phase type if the PV generator voltage is controlled [6]. Shezan et al. [7], this paper proposed wind and solar hybrid renewable energy system and simulated on MATLAB. Fuzzy and Logic MPPT control is used for improve the system efficiency. This system is efficient for both AC and DC load in remote areas. Simulation model of IPS with energy management controller is developed using MATLAB/ Simulink R2011b. This system is suitable for telecom equipment where constant voltage and continuous power is required. This system maintains the 48 V DC and three phase AC supply [7]. Kanagasakthivel et al. [8], this paper presents the hybrid model of PV array solar energy and induction generator driven wind energy. They simulated individually first then hybrid the PV and Wind system under different load. They plot the graph of wind speed characteristics, rotor speed in wind system, torque in wind system and output voltage. In PV solar module, they used 18 solar cells are connected in series to get Voltage 36 V and current 4.6 A. They used irradiation G = 0.8 W/m2 , wind speed 14 m/s and Load = 500 Ω in hybrid system. This system is useful of remote areas and simulated on MATLAB/SIMULINK [8]. Kumar et al. [9], this paper used Combined Modified Bat Search algorithm and artificial neural network control system (CMBSNN) for Hybrid Renewable Energy System (HRES). CMBSNN algorithm optimize the reactive power according to controller parameters. Bat algorithm increase searching behaviour and utilizing the smart mathematical function. It uses proportional integral (PI) controller which is reduce the system error. Artificial Neural Network is used to machine learning. They simulate the whole system on MATLAB Simulink and plot the graph with various parameters [9]. Phan and Lai [10], this paper represent the Hybrid Renewable energy system using Reinforcement Learning Approach (RLA) for microgrid. They used deep Q-learning method for power management control system. They proposed the modified HRES and give a block diagram. In block diagram they connect PV, Battery, Generator, fuel cell and Hydrogen tank with Ac/Dc Inverter. They selected location at BASCO Island in Philippines. At present it uses diesel generator system only. So, government want to introduce low-cost environment friendly system. The proposed system fulfils the all demand and criteria of Basco Island. They also give the environment data of Basco Island. They modified the MPPT system and combine the Q-learning and P&O

286

H. Giroh et al.

method to overcome the disadvantages of both systems. This h-POQL method flow chart diagram is shown and explain. They plot the various graph of performance of system on different weather condition and compare with other systems. They also want to implement physically on Basco Island in future [10]. Kumar et al. [11], they published this paper on topic Design and analysis of Hybrid RBFN MPPT solar and wind energy system. They proposed a single MPPT topology for hybrid system and give in block diagram. They implement basic PV diode current, voltage and power equation and give the all-parameter rating of proposed system. They simulate the Simulink Model on MATLAB and integrated with 230 V, 50 Hz AC grid. They plot graph of output voltage, current and power of stand-alone system as well as gird connected system. All the results summarise in table under active power and reactive power column. They reduce the complexity of hybrid system successfully [11]. Habib et al. [12], they design a standalone hybrid renewable energy system for small town of Pakistan. They proposed the advance Power Management System (PMS) and design & simulate on MATLAB/Simulink software. Their model consists of wind generation system, Battery storage, Battery control unit, Inverter, Filter, Model predicative control Block and Load. They optimize the size of individual components, input the real data of particular place and get output and validate with conventional model results. This system is able to fulfil the demand of power supply of remote area of Pakistan with excess energy of 30%. TDH reduced to 0.26%. It saves the 76.7% cost with 100% clean energy [12]. Baumann et al. [13], they published the paper on the topic “Energy Flow Management of hybrid renewable energy system”. They set up an experiment at Ostfalia, Germany. It consists of 5.1 k Wp PV-array, a 1 k Wp PV-array with tilt angle, a 4-kW micro wind turbine, a 1.2 kW fuel cell system, and a 6 kW micro-CHP unit. They give block diagram of whole system at laboratory which connected with LON and web server. It stores the all data into MySql Database and driven from MATLAB Simulink. They plot the graph of total electrical power of whole day and 2-h duration. They also give the flow chart diagram of energy management algorithm. This system produces 164.03 Kwh but need 72 Kwh so it generates surplus power. They physically implemented the first simple model successfully and going to implement bidirectional complex energy management system in future [13]. Saidi and Chellali [14], they published the paper on simulation and control hybrid renewable power. In introduction section described the components of hybrid system. PV system have PV array, MPPT, boost converter, wind fan, wind turbine, PMSG and control system. They explain the PV array system and give the equations of I-V and P–V and sketch the graph. They proposed modified hybrid system as given in block diagram. They implement Fuzzy Logic Controller on subsystem error generator and supply to inverter voltage line then calculate output. They show the all the output voltage of inverter in graph and compared. Their FLC based model is more efficient and reliable than PI voltage regulated inverter [14]. Qiang et al. [15], they introduced multi directional power converter in Hybrid renewable energy. They used high frequency transformer for distributed generation

To Analyse the Impact of Integration of Wind and Solar Power …

287

hybrid system. In proposed schematic diagram, wind energy and solar energy integrated with first and second input stages, output and storage circuit stages, Inverter and storage devices. The proposed four direction converter worked in 10 possible modes and given by wind (W), PV (P), storage (S) and Load (L) combination. All equivalent circuit of converter is represented. Simulate the equivalent circuit on Pspice and MATLAB with given parameters and output current and voltage graph is given as result [15]. Anoune et al. [16], they published their paper on topic Maximize the production of electrical power from Hybrid Renewable energy. They stated that alone PV or Wind energy system cannot fulfil the demand of local demand. So, hybridization of PV and wind is compulsory. It reduces the both disadvantages. The proposed a model consists of MPPT, DC/DC Boost converter and Lead acid battery as storage device. They simulate their proposed DC-DC boost converter model on Simulink and show the output voltage stable on 48 V DC. P&O method is used to tack maximum point of power. They have used PWM techniques for control the DC-DC boost converter. Overall model is stable and reliable [16]. Zahboune et al. [17], they optimize Hybrid Renewable energy source and Modified Electric System Cascade Analysis with the help of Homer software. Homer software used for design the best combination of renewable energy sources i.e., solar, wind, hydroelectricity, bio-diesel generator. They also used MESCA method where collect the monthly record of PV, wind and charging-discharging history of battery in table. In first algorithm wind turbine and batteries number are calculated as FEE at T = 288 h but second algorithm sizing correction at T = 8760 h. The homer result is given as graph of electrical and leaking energy. They also compared the Homer pro and MESCA method and tables & graphs are showed. In future, MESCA method can be applied on green energy and storage system [17]. Mokhtara et al. [18], they focused on supply demand management of Hybrid renewable energy systems for residential area. They used PSO algorithm to reduced building energy consumption. In this paper, they described system modelling and configuration of HRES. For this modelling, they draw a schematic diagram of HRES System and estimate the average residential load. They calculate the supply power of PV energy, wind energy and diesel generator and demand from household equipment of each room of house. After simulation of proposed model on MATLAB And HOMER software, they find out results and compared to each other for validation and discussed [18].

4 Hybrid Renewable Energy System for Different Geographical Condition A hybrid renewable energy system combines multiple sources of renewable energy to provide a reliable and efficient energy supply. The specific components of the system will depend on the geographical conditions and energy demands of the location. For

288

H. Giroh et al.

example, in areas with strong wind and solar resources, a hybrid system may include wind turbines and photovoltaic (PV) panels. In regions with more consistent hydro resources, a system may include hydro power and wind or solar. The system can also be designed to incorporate energy storage options, such as batteries, to smooth out fluctuations in renewable energy production. The goal of a hybrid renewable energy system is to maximize the use of locally available renewable resources, reduce dependence on fossil fuels, and provide a cost-effective and sustainable energy solution [19].

4.1 Status of Hybrid System Power Generation System for Uttarakhand, Haryana, and Rajasthan Wind and solar power generation systems would be well-suited for the states of Uttarakhand, Haryana, and Rajasthan. These states have abundant sunlight and wind resources, making them ideal for the deployment of photovoltaic (PV) panels and wind turbines [20]. For Uttarakhand, a hybrid system incorporating wind and solar would be ideal, as the state has a good wind energy potential in some areas and good solar potential in others. Wind turbines could be deployed in areas with strong wind resources, while PV panels could be installed in areas with high solar insolation. In Haryana, a similar hybrid system could be deployed, but with a greater emphasis on solar power. Haryana is known for its high solar insolation, making it a prime location for the deployment of large-scale PV panels. In Rajasthan, a wind-solar hybrid system could also be deployed, taking advantage of the strong wind resources in some areas and the high solar insolation in others. In this state, wind turbines could be deployed in areas with high wind speeds, while PV panels could be installed in areas with high solar insolation. In all cases, energy storage systems, such as batteries, could be incorporated into the hybrid systems to ensure that energy is available even when renewable energy production is low. This would help to provide a more stable and reliable energy supply for the region [21].

4.2 Analyse the Wind and Solar Power Generation System for Uttarakhand, Haryana, and Rajasthan The analysis of wind and solar power generation systems in Uttarakhand, Haryana, and Rajasthan would consider the following factors: • Renewable energy resources: The first step is to assess the wind and solar resources in each state. This can be done using wind and solar data from meteorological stations [22], satellite imagery, and numerical models. The data can be used to determine the wind speeds, solar insolation, and other relevant parameters [11].

To Analyse the Impact of Integration of Wind and Solar Power …

289

Table 1 Wind and solar energy in Uttarakhand, Haryana, and Rajasthan State

Average wind speed (m/s)

Average solar insolation (kWh/m2 /day)

Uttarakhand

6–8

5–6

Haryana

5–7

6–7

Rajasthan

6–8

6–7

• Energy demand: The energy demand of each state must be estimated to determine the scale of the renewable energy system required. This includes the electricity demand of households, businesses, and industries in each state [23]. • Techno-economic feasibility: The cost of installing and operating wind and solar power generation systems must be evaluated. This includes the cost of the wind turbines or PV panels, the cost of energy storage systems, and the operational and maintenance costs [24]. • Grid integration: The renewable energy system must be integrated into the existing grid infrastructure. This involves ensuring that the energy produced by the wind and solar systems can be delivered to the grid and consumed by end-users [25]. • Environmental impact: The environmental impact of the wind and solar power generation systems must be evaluated. This includes the impact of the wind turbines or PV panels on the local ecology and wildlife, as well as the impact of the energy production and transmission processes [26]. Based on the results of this analysis, a wind and solar power generation system can be designed and deployed in each state to meet the energy demands of the region while taking into account the local conditions, resources, and constraints (Table 1). The above reflect the actual wind and solar energy conditions in each state. The data may vary based on the specific location and the time of year.

5 Scope of Research and Adoptability of Machine Learning Regional studies of solar and other renewable energy resource have been widely covered in many research. The regional context helps to industry to plan and implement their set up with consideration of the underlying research. In our research we also initiated with demography analysis and adaptability of renewable energy. Beside the Uttarakhand, Haryana, and Rajasthan together in context of hybrid renewable energy has been still untouched and open research. This analysis gives the adaptability of alternative energy in norther region of India. The scope of research on the integration of wind and solar power generation systems in Uttarakhand, Haryana, and Rajasthan using machine learning (ML) techniques is broad and promising. ML algorithms can be applied to various aspects of this integration, including resource assessment, power generation forecasting, grid stability analysis, energy management optimization, and environmental impact assessment. ML can help in assessing

290

H. Giroh et al.

the potential of wind and solar resources by analysing historical weather patterns, topography, and other relevant data. It can accurately forecast power generation by considering real-time weather data, historical generation patterns, and other variables. ML models can simulate and analyze the impact of integrating renewable energy sources into the existing grid infrastructure, ensuring stability and reliability. The adoptability of machine learning in this research is high. ML techniques have been successfully employed in various domains and have proven effective in analyzing complex and large datasets. With the availability of historical data, realtime monitoring systems, and advancements in ML algorithms, the integration of wind and solar power generation systems can benefit greatly from ML’s ability to analyze, predict, and optimize various aspects of the system [28].

6 Conclusion and Future Scope The wind and solar energy have great potential in Uttarakhand, Haryana, and Rajasthan. These states are blessed with abundant wind and solar resources, making them ideal locations for the development of renewable energy. With the right policy framework and investment, wind and solar energy can play a significant role in meeting the energy demands of these states and reducing the dependence on conventional energy sources. There is a huge scope for the future growth of wind and solar energy in these states. The government and private sector can work together to improve infrastructure and grid connectivity, attract investment, and create job opportunities in the renewable energy sector. Additionally, advances in technology and the falling cost of renewable energy systems will also help to increase the adoption of wind and solar energy in these states. In conclusion, wind and solar energy hold great promise for Uttarakhand, Haryana, and Rajasthan and the future looks bright for the renewable energy sector in these states. Future the integration of wind and solar power generation systems in Uttarakhand, Haryana, and Rajasthan can benefit from the application of machine learning (ML) techniques. ML algorithms could effectively analyse data related to resource potential, power generation forecasting, grid integration, energy management optimization, and environmental impact assessment. The insights generated through ML-based analysis provide valuable information for policymakers and stakeholders, enabling informed decision-making and supporting the transition towards sustainable and clean energy sources. By leveraging the capabilities of machine learning, we can accelerate the adoption of renewable energy and contribute to a greener and more resilient energy infrastructure in these regions.

To Analyse the Impact of Integration of Wind and Solar Power …

291

References 1. Herbert, G. J., Iniyan, S., Sreevalsan, E., & Rajapandian, S. (2007). A review of wind energy technologies. Renewable and sustainable energy Reviews, 11(6), 1117–1145. 2. Owusu, P. A., & Asumadu-Sarkodie, S. (2016). A review of renewable energy sources, sustainability issues and climate change mitigation. Cogent Engineering, 3(1), 1167990. 3. Poudyal, R., Loskot, P., Nepal, R., Parajuli, R., & Khadka, S. K. (2019). Mitigating the current energy crisis in Nepal with renewable energy sources. Renewable and Sustainable Energy Reviews, 116, 109388. 4. Home Based Internship Certificate. (n.d.). Retrieved April 4, 2023, from https://www.ametuniv. ac.in/DVV_Criteria_1/134_FieldProjects_Internship_2019_20/Aswin_Renjith.pdf 5. Amer, M., Namaane, A., & M’sirdi, N. K. (2013). Optimization of hybrid renewable energy systems (HRES) using PSO for cost reduction. Energy Procedia, 42, 318–327. 6. Andoulssi, R., Draou, A., Jerbi, H., Alghonamy, A., & Khiari, B. (2013). Nonlinear control of a photovoltaic water pumping system. Energy Procedia, 42, 328–336. 7. Shezan, S., Khan, N. H., Anowar, M. T., Delwar, M. H., Islam, M. D., Reduanul, M. H., & Kabir, M. A. (2016). Fuzzy logic implementation with MATLAB for solar-wind-battery-diesel hybrid energy system. Imperial Journal of Interdisciplinary Research (IJIR), 2(5), 574–583. 8. Kanagasakthivel, B., & Devaraj, D. (2015, February). Simulation and performance analysis of Solar PV-Wind hybrid energy system using MATLAB/SIMULINK. In 2015 international conference on computing and communications technologies (ICCCT) (pp. 99–104). IEEE. 9. Desabathina, N.V.M., Merugu, S., Gunjan, V.K., Kumar, B.S. (2022). Agricultural crowdfunding through blockchain. In: Kumar, A., Senatore, S., Gunjan, V.K. (Eds) ICDSMLA 2020. Lecture notes in electrical engineering (Vol. 783). Singapore: Springer. https://doi.org/10.1007/ 978-981-16-3690-5_155 10. Phan, B. C., & Lai, Y. C. (2019). Control strategy of a hybrid renewable energy system based on reinforcement learning approach for an isolated microgrid. Applied Sciences, 9(19), 4001. 11. Panwar, D., Saini, G. L., & Agarwal, P. (2022). Human eye vision algorithm (HEVA): A novel approach for the optimization of combinatorial problems. In: Garg, L., Basterrech, S., Banerjee, C., Sharma, T., & K. (Eds) Artificial intelligence in healthcare. Advanced technologies and societal change. Singapore: Springer. https://doi.org/10.1007/978-981-16-6265-2_5 12. Balodi, A., Mangla, N., Hombalimath, A., & Manjula, H. T. (2022). Analytics and big data in the health domain. In: Garg, L., Basterrech, S., Banerjee, C., & Sharma, T. K. (Eds.), Artificial intelligence in healthcare. Advanced technologies and societal change. Singapore: Springer. https://doi.org/10.1007/978-981-16-6265-2_6 13. K, L. P., Surya Narayana, G., Ansari, M. D., & Gunjan, V. K. (2021). Instantaneous approach for evaluating the initial centers in the agricultural databases using K-means clustering algorithm. Journal of Mobile Multimedia, 18(1), 43–60. https://doi.org/10.13052/jmm1550-4646. 1813 14. Saidi, A., & Chellali, B. (2017, May). Simulation and control of solar wind hybrid renewable power system. In 2017 6th international conference on systems and control (ICSC) (pp. 51–56). IEEE. 15. Noor-e-Alam Siddiquee, K., Shabiul Islam, M.d., Singh, N., Gunjan, V. K., Yong, W. H., Huda, M. N., Bhupal Naik, D. S. (2022). Development of algorithms for an IoT-based smart agriculture monitoring system. Wireless Communications and Mobile Computing, 2022, Article ID 7372053, 16. https://doi.org/10.1155/2022/7372053 16. Anoune, K., Bouya, M., Ghazouani, M., Astito, A., & Abdellah, A. B. (2016, November). Hybrid renewable energy system to maximize the electrical power production. In 2016 international renewable and sustainable energy conference (IRSEC) (pp. 533–539). IEEE. 17. Zahboune, H., Zouggar, S., Krajacic, G., Varbanov, P. S., Elhafyani, M., & Ziani, E. (2016). Optimal hybrid renewable energy design in autonomous system using modified electric system cascade analysis and homer software. Energy Conversion and Management, 126, 909–922.

292

H. Giroh et al.

18. Mokhtara, C., Negrou, B., Bouferrouk, A., Yao, Y., Settou, N., & Ramadan, M. (2020). Integrated supply–demand energy management for optimal design of off-grid hybrid renewable energy systems for residential electrification in arid climates. Energy Conversion and Management, 221, 113192. 19. Ahmed, S. M., Kovela, B., & Gunjan, V. K. (2021). Solar-powered smart agriculture and irrigation monitoring/control system over cloud—An efficient and eco-friendly method for effective crop production by farmers in rural India. In Gunjan, V. K., & Zurada, J. M. (Eds.), Proceedings of international conference on recent trends in machine learning, IoT, smart cities and applications. Advances in intelligent systems and computing (Vol. 1245). Singapore: Springer. https://doi.org/10.1007/978-981-15-7234-0_24 20. Bhowmik, T., Majumdar, S., Choudhury, A., Banerjee, A., & Roy, B. (2023). Importance of internal and external psychological factors in digital learning. In: Choudhury, A., Biswas, A., & Chakraborti, S. (Eds.), Digital learning based education. Advanced technologies and societal change. Springer, Singapore. https://doi.org/10.1007/978-981-19-8967-4_8 21. Das, N. (2023). Digital education as an integral part of a smart and intelligent city: A short review. In Choudhury, A., Biswas, A., & Chakraborti, S. (Eds.), Digital learning based education. Advanced technologies and societal change. Singapore: Springer. https://doi.org/10.1007/ 978-981-19-8967-4_5 22. Kirar, A., Bhalerao, S., Verma, O. P., & Ansari, I. A. (2022). Protecting ECG signals with hybrid swarm intelligence algorithm. In: Garg, L., Basterrech, S., Banerjee, C., & Sharma, T. K. (Eds.), Artificial intelligence in healthcare. Advanced technologies and societal change. Singapore: Springer. https://doi.org/10.1007/978-981-16-6265-2_4 23. Reddy, S., & Painuly, J. P. (2004). Diffusion of renewable energy technologies—barriers and stakeholders’ perspectives. Renewable Energy, 29(9), 1431–1447. 24. Mamaghani, A. H., Escandon, S. A. A., Najafi, B., Shirazi, A., & Rinaldi, F. (2016). Technoeconomic feasibility of photovoltaic, wind, diesel and hybrid electrification systems for off-grid rural electrification in Colombia. Renewable Energy, 97, 293–305. 25. Gunjan, V. K., Kumar, S., Ansari, M. D., & Vijayalata, Y. (2022). Prediction of agriculture yields using machine learning algorithms. In: Gunjan, V. K., & Zurada, J. M. (Eds.), Proceedings of the 2nd international conference on recent trends in machine learning, IoT, smart cities and applications. Lecture notes in networks and systems (Vol. 237). Singapore: Springer. https:// doi.org/10.1007/978-981-16-6407-6_2 26. Jaber, S. (2013). Environmental impacts of wind energy. Journal of Clean Energy Technologies, 1(3), 251–254. 27. Mosavi, A., & Varkonyi-Koczy, A. R. (2017). Integration of machine learning and optimization for robot learning. In Recent global research and education: technological challenges: proceedings of the 15th international conference on global research and education inter-academia 2016 (pp. 349–355). Springer International Publishing. 28. Syafrudin, M., Alfian, G., Fitriyani, N. L., & Rhee, J. (2018). Performance analysis of IoTbased sensor, big data processing, and machine learning model for real-time monitoring system in automotive manufacturing. Sensors, 18(9), 2946.

VLSI Implementation of an 8051 Microcontroller Using VHDL and Re-Corrective Measure Using AI Tushar Vardhan Zen

1 Introduction A microcontroller is a small computer on a single integrated circuit. It contains a processor, memory, and input/output peripherals [1]. The 8051 microcontroller is a widely used microcontroller in embedded systems. It was introduced in the 1980s by Intel and has since been used in various applications such as automotive, industrial, and consumer electronics [2]. VHDL is a hardware description language that allows for the creation of digital circuits and systems. The objective of this research paper is to implement an 8051-microcontroller using VHDL. The implementation will include the design of the processor, memory, and input/output peripherals. The implementation of 8051 microcontroller using VHDL will allow for the creation of customized microcontrollers with specific functionalities [3].

2 Background The 8051 microcontroller is a popular 8-bit microcontroller architecture that has been widely used in various applications since its introduction in 1980 [4]. Its simplicity, low cost, and ease of use have made it a popular choice for many embedded systems. VHDL is a hardware description language used for designing digital circuits and systems. It is often used to design and simulate digital circuits before implementation on an FPGA or ASIC [5]. Implementing the 8051-microcontroller using VHDL involves designing a digital circuit that can perform the same functions as the physical 8051 microcontroller. The VHDL design must include the CPU, memory, I/O ports, T. V. Zen (B) Electronics and Communication Engineering, New Delhi, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 V. K. Gunjan et al. (eds.), Modern Approaches in Machine Learning and Cognitive Science: A Walkthrough, Studies in Computational Intelligence 1117, https://doi.org/10.1007/978-3-031-43009-1_24

293

294

T. V. Zen

timers, and other peripherals found on the physical 8051 microcontroller [6]. The VHDL design should also include a way to interface with external devices such as sensors, actuators, and communication interfaces. The VHDL design of the 8051 microcontrollers can be done using different design methodologies such as a topdown approach or a bottom-up approach. In a top-down approach, the designer starts by defining the system requirements and then develops a high-level description of the system. The high-level description is then refined to a lower level until the final implementation is reached. In a bottom-up approach, the designer starts by developing individual components and then integrates them into the final design. The 8051 microcontroller is a popular 8-bit microcontroller that was first introduced by Intel in 1980. It is widely used in various applications, such as automotive systems, industrial automation, and consumer electronics. VHDL is a hardware description language used to model digital circuits and systems. The combination of the 8051 microcontrollers with VHDL can lead to powerful and efficient systems [7]. The implementation of the 8051 microcontrollers with VHDL involves creating a hardware description of the microcontroller in VHDL. This is done by specifying the various components of the microcontroller, such as the registers, ALU, and instruction decoder, in VHDL code. The VHDL code is then synthesized to generate a netlist, which is a description of the circuit in terms of gates and flip-flops. The VHDL code for the 8051 microcontrollers can be developed using various tools, such as Xilinx ISE [8]. These tools provide a graphical user interface that allows designers to create and edit the VHDL code, simulate the design, and synthesize the netlist. Once the netlist is generated, it can be programmed onto a field-programmable gate array (FPGA) or an application-specific integrated circuit (ASIC). The FPGA or ASIC will then function as the 8051 microcontrollers, executing the instructions specified in the VHDL code. The benefits of implementing the 8051 microcontrollers with VHDL are numerous. First, VHDL allows for easy simulation and verification of the microcontroller design, which helps in identifying and correcting errors before the design is implemented in hardware. Second, VHDL enables the integration of the 8051 microcontrollers with other digital systems, such as communication interfaces and peripheral devices. Finally, VHDL facilitates the development of complex systems that require the 8051 microcontrollers to interface with various digital systems (Fig. 1).

3 Proposed Architecture and Methodology CPU Module The CPU module of the 8051 microcontroller is responsible for executing the instruction and control the overall operation. The VHDL code for the CPU module includes the ALU, register file, program counter, and control unit. The ALU performs arithmetic and logical operations, while the register file stores the data and addresses. The program counter stores the address of the next instruction to be executed, and the control unit generates the control signals to execute the instruction [9].

VLSI Implementation of an 8051 Microcontroller Using VHDL and …

295

Fig. 1 Block diagram of 8051 microcontroller

Memory Module The memory module of the 8051 microcontroller consists of program memory (ROM) and data memory (RAM). The program memory stores the program code and is non-volatile, while the data memory stores the data and is volatile. The VHDL code for the memory module includes the ROM, RAM, and memory interface. The memory interface provides the data and address signals to access the memory [10]. I/O Ports Module Input/output (I/O) ports: The 8051 microcontroller has four 8-bit I/O ports (P0, P1, P2, and P3) that can be used to interface with external devices. These ports can be configured as inputs or outputs depending on the application. The VHDL code for the I/O ports module includes the input/output buffers, latch, and control logic [11]. Timers Module The 8051 microcontroller has two 16-bit timers/counters (T0 and T1) that can be used for a variety of tasks, including timing events and counting external events. The VHDL code for the timers module includes the timer/counter, control logic, and interrupt generator [12]. Interrupts Module The 8051 microcontroller has five interrupt sources that can be used to interrupt the normal program flow and perform a specific task. These include two external

296

T. V. Zen

interrupts (INT0 and INT1), two timer interrupts (T0 and T1), and a serial port interrupt (RI/TI). The VHDL code for the interrupts module includes the interrupt controller and priority encoder [12].

4 Implementation and Result Designing and implementing the VLSI implementation of the 8051 microcontrollers with VHDL involves several steps. The process can be broadly divided into two stages: designing the microcontroller architecture in VHDL and implementing the design on an FPGA. Designing the Microcontroller Architecture in VHDL: Requirement Analysis: The first step in designing the microcontroller architecture is to identify the requirements of the system. This involves determining the input and output requirements, memory requirements, and the instruction set of the microcontroller. Design Specification: Based on the requirements, the design specification of the microcontroller is created. This specification includes the functional blocks, such as the instruction decoder, register bank, arithmetic logic unit (ALU), and interrupt controller as shown in Fig. 2. VHDL Coding: In this part, 8051 microcontroller architecture is coded in VHDL. This involves writing VHDL code for the individual functional blocks and integrating them to create the complete microcontroller design as shown in Fig. 3. Fig. 2 Design specification diagram

VLSI Implementation of an 8051 Microcontroller Using VHDL and …

297

Fig. 3 VHDL file structure

Figure 3 shows that a top module named I8051_All is used structure modelling to connect all the peripheral module ALU, RAM, ROM etc. all other module is used behavioural modelling to implement ALU, RAM, ROM, control unit, timers and interrupt [13]. Simulation: Once the VHDL code is written, we are using ISE simulator to verify its correctness and functionality. The simulation results are then analysed and any necessary changes are made to the VHDL code as shown in Fig. 4. Figure 4 shows that simulation of top module. Which take 8 bit data each from p0_in to p3_in and give output 8 bit data also on each p0_out to p3_out. Synthesis: After simulation, the VHDL code is synthesized to generate a netlist that can be implemented on an FPGA or ASIC. This involves mapping the VHDL code

Fig. 4 Simulation of 8051_ALL entity on ISE Simulator

298

T. V. Zen

to FPGA or ASIC primitives and optimizing the design for area, power, and timing as shown in Figs. 5, 6, 7, 8 and 9. Figure 5 clearly shows that we synthesize the proposed 8051 microcontroller on Xilinx 8.1 successfully and we used 68% of LUTs 79% of Slices utilization available in xcv150-6bg256 device. Figure 6 shows that RTL schematic diagram of top module of proposed 8051 microcontroller. In Fig. 7 all the chip level circuit internal diagram of top module is shown. It also clearly shows that how to connect the other peripheral devices of 8051 to each other. Figure 8 shows that optimization the design by adjusting synthesis settings to achieve better performance, area utilization, and timing.

Fig. 5 Device utilization summary of 8051 microcontroller project on Xilinx 8.1 Fig. 6 RTL schematic diagram of top module

VLSI Implementation of an 8051 Microcontroller Using VHDL and …

Fig. 7 RTL diagram of all chip level diagram

Fig. 8 FPGA level diagram and routing Fig. 9 Voltage current and power of proposed 8051 microcontroller

299

300

T. V. Zen

Figure 9 shows that Voltage current and power of proposed 8051 microcontroller. We achieve 26.79 mW power dissipation on working condition of 8051 microprocessor. Implement the Netlist on an FPGA: Generate a programming file (.bit) from Xilinx ISE project. This can be done by selecting “Generate Programming File” from the “Processes” menu in the Xilinx ISE Project Navigator. Open iMPACT and create a new project. Select the appropriate device family and device for our project. Add the programming file (.bit) to ouriMPACT project. Selecting “Add File” from the “Project” menu, and then selecting the bit file. Connect your programming cable to the target FPGA device and computer. Configure the programming properties in iMPACT. This includes selecting the programming cable, the target device, and the programming file (.bit). Click the “Program” button in iMPACT to program our target device with the.bit file. Verify that the programming was successful by testing our FPGA device and checking that it behaves as expected.

5 Corrective Measure Using AI in VLSI Implementation of an 8051-Microcontroller In the VLSI implementation of an 8051-microcontroller using VHDL, AI can be utilized to provide re-correction measures during the design and implementation process. Below how AI can be applied at different stages to enhance the design: Design Optimization: AI algorithms, such as genetic algorithms [14–18] or reinforcement learning, can optimize the microcontroller’s design parameters. By iterating through multiple design variations and evaluating their performance, AI can identify areas where improvements can be made. This could include optimizing component placement, routing, or logic design to achieve better performance, power consumption, or area utilization. Bug Detection and Correction: AI-based static analysis or formal verification techniques can help detect potential design bugs or flaws. These algorithms can analyze the VHDL code, identify common coding errors or design violations, and suggest corrective measures. This can help eliminate design issues before they manifest into more significant problems during the implementation phase. Performance Prediction and Analysis: AI models can be trained on large datasets of VHDL designs and their corresponding performance metrics. By analyzing the microcontroller’s VHDL code and input/output patterns, AI models can predict performance metrics such as power consumption, timing, or area utilization. These predictions can assist in identifying potential bottlenecks or areas of improvement, allowing designers to make informed decisions during the design process.

VLSI Implementation of an 8051 Microcontroller Using VHDL and …

301

Intelligent Place and Route: AI algorithms can be used to optimize the placement and routing of the microcontroller’s components on the chip. By considering factors like signal integrity, wirelength, and performance goals, AI-based place and route techniques can suggest improved placement and routing strategies. This can lead to reduced signal delays, improved timing closure, and better overall performance. Automated Testing and Debugging: AI can be used to automate the generation of test stimuli and the analysis of simulation results. By leveraging machine learning or symbolic reasoning techniques, AI algorithms can generate comprehensive test scenarios, identify critical paths, and detect potential functional or timing issues. This automated testing process can help accelerate the verification process and improve the reliability of the microcontroller design. Adaptive Optimization: AI algorithms can continuously monitor the performance of the implemented microcontroller and dynamically optimize its operation. By analysing runtime data and considering varying environmental conditions, AI can adapt the microcontroller’s configuration or operating parameters to maximize performance, minimize power consumption, or improve reliability [19, 20].

6 Conclusion and Future Scope The implementation of 8051 microcontroller with VHDL offers a powerful solution for designing and simulating complex digital circuits. The 8051 microcontroller is a widely used and versatile microcontroller with a rich instruction set and support for various peripherals. By using VHDL, a hardware description language, designers can create efficient and highly optimized implementations of the 8051 microcontrollers for a wide range of applications. The implementation of 8051 microcontroller with VHDL offers several benefits, including reduced development time, increased performance, and reduced power consumption. VHDL allows designers to create highly optimized digital circuits that can be easily modified and reused for future designs. The 8051 microcontrollers, with its extensive instruction set and support for various peripherals, provides a powerful platform for developing embedded systems. Overall, the implementation of 8051 microcontroller with VHDL offers a powerful solution for designing and simulating complex digital circuits, and the future looks bright for this technology as it continues to evolve and meet the needs of the industry. Further the VLSI implementation of an 8051-microcontroller using VHDL and the integration of AI techniques offer significant benefits and opportunities for improvement. By leveraging AI algorithms throughout the design flow, designers can achieve enhanced performance, reduced power consumption, and improved design quality. AI can optimize the microcontroller’s design parameters, detect and correct bugs, predict performance metrics, optimize component placement and routing, automate testing and debugging, and enable adaptive optimization. The future scope of this integration includes exploring deep learning for design automation, reinforcement learning

302

T. V. Zen

for adaptive optimization, hardware-software co-design, fault detection and tolerance, high-level synthesis, performance and power analysis, and hardware security. Continued research and development in these areas will contribute to the advancement of microcontroller designs, enabling more efficient and intelligent embedded systems.

References 1. Swetha, B., Blesington, T. P., Basha, F. N., & BhanuMurthy, B. (2013). Design of 32-bit microcontroller processor in soc. International Journal of Advances in Engineering and Technology, 6(2), 812. 2. MacKenzie, I. S. (2007). The 8051 microcontroller. Pearson Education India. 3. Rudra Kumar, M., Pathak, R., & Gunjan, V. K. (2022). Machine learning-based project resource allocation fitment analysis system (ML-PRAFS). In Kumar, A., Zurada, J. M., Gunjan, V. K., Balasubramanian, R. (Eds.), Computational intelligence in machine learning. Lecture notes in electrical engineering (Vol. 834). Singapore: Springer. https://doi.org/10.1007/978-981-168484-5_1 4. Çakiroglu, M. (2010). Software implementation and performance comparison of popular block ciphers on 8-bit low-cost microcontroller. International Journal Physics Science, 5(9), 1338– 1343. 5. Amara, A., Amiel, F., & Ea, T. (2006). FPGA versus ASIC for low power applications. Microelectronics journal, 37(8), 669–677. 6. Gaddam, D. K. R., Ansari, M. D., Vuppala, S., Gunjan, V. K., & Sati, M. M. (2022). A performance comparison of optimization algorithms on a generated dataset. In Kumar, A., Senatore, S., & Gunjan, V. K. (Eds.), ICDSMLA 2020. Lecture notes in electrical engineering (Vol. 783). Singapore: Springer. https://doi.org/10.1007/978-981-16-3690-5_135 7. Verma, L., Pottinger, H. J., & Beetner, D. G. (2003, June). A software debugger interface for an 8051 hardware model. In Proceedings 2003 IEEE international conference on microelectronic systems education. MSE’03 (pp. 112–114). IEEE. 8. Rodriguez-Andina, J. J., Moure, M. J., & Valdes, M. D. (2007). Features, design tools, and application domains of FPGAs. IEEE Transactions on Industrial Electronics, 54(4), 1810– 1823. 9. Navalgund, S. S., & Tonse, P. R. (2013). Design, Development and Implementation of ALU, RAM and ROM for 8051 Microcontroller on FPGA using VHDL. International Journal of Computer Applications, 975, 8887. 10. Cardarilli, G. C., Ottavi, M., Pontarelli, S., Re, M., & Salsano, A. (2005). Fault tolerant solid state mass memory for space applications. IEEE Transactions on Aerospace and Electronic Systems, 41(4), 1353–1372. 11. Siddiquee, K. N. E. A., Islam, M. S., Singh, N., Gunjan, V. K., Yong, W. H., Huda, M. N., & Naik, D. B. (2022). Development of algorithms for an IoT-based smart agriculture monitoring system. Wireless Communications and Mobile Computing, 2022, Article ID 7372053, 16. https://doi. org/10.1155/2022/7372053 12. Qian, K., Den Haring, D., Cao, L., Qian, K., den Haring, D., & Cao, L. (2009). 8051 microcontroller. Embedded Software Development with C, 73–96. 13. Singh, N., Gunjan, V. K., & Nasralla, M. M. (2022). A parametrized comparative analysis of performance between proposed adaptive and personalized tutoring system “seis tutor” with existing online tutoring system. IEEE Access, 10, 39376–39386. https://doi.org/10.1109/ACC ESS.2022.3166261 14. Kramer, O., & Kramer, O. (2017). Genetic algorithms (pp. 11–19). Springer International Publishing.

VLSI Implementation of an 8051 Microcontroller Using VHDL and …

303

15. Raval, M., Bhardwaj, S., Aravelli, A., Dofe, J., & Gohel, H. (2021). Smart energy optimization for massive IoT using artificial intelligence. Internet of Things, 13, 100354. 16. Ahmed, M., Ansari, M. D., Singh, N., Gunjan, V. K., Santhosh Krishna B. V., & Khan, M. (2022). Rating-based recommender system based on textual reviews using IoT smart devices. Mobile Information Systems, 2022, Article ID 2854741, 18. https://doi.org/10.1155/2022/285 4741 17. Bhowmik, T., Majumdar, S., Choudhury, A., Banerjee, A., & Roy, B. (2023). Importance of internal and external psychological factors in digital learning. In: Choudhury, A., Biswas, A., Chakraborti, S. (Eds.), Digital learning based education. Advanced technologies and societal change. Singapore. Springer. https://doi.org/10.1007/978-981-19-8967-4_8 18. Das, N. (2023). Digital education as an integral part of a smart and intelligent city: A short review. In: Choudhury, A., Biswas, A., Chakraborti, S. (Eds.), Digital learning based education. Advanced technologies and societal change. Singapore: Springer. https://doi.org/10.1007/978981-19-8967-4_5 19. Karthik, R., Shukla, P., Lavanya, S., Naga Satish, L. L., & Sai Rohith Krishna, J. (2022). Deep transfer learning for detecting cyber attacks. In: Garcia Diaz, V., & Rincón Aponte, G. J. (Eds.), Confidential computing. Advanced technologies and societal change. Singapore: Springer. https://doi.org/10.1007/978-981-19-3045-4_12 20. Guru Akhil, T., Pranay Krishna, Y., Gangireddy, C., Kumar, A. K., & Sowjanya, K. L. (2022). Cyber-hacking breaches for demonstrating and forecasting. In: Garcia Diaz, V., & Rincón Aponte, G. J. (Eds.), Confidential computing. Advanced technologies and societal change. Singapore: Springer. https://doi.org/10.1007/978-981-19-3045-4_17

K-Mean Energy Efficient Optimal Cluster Based Routing Protocol in Vehicular Ad-Hoc Networks A. C. Pise and K. J. Karande

1 Introduction VANET features are most comparable to those of mobile ad hoc networks (MANETs) [3, 4], which shows that both types of networks are self-organizing and selfmanaging, have low bandwidth, and remain in the same area while sharing radio transmission. As a result of the greater speed and mobility of VANET nodes in comparison to MANETs, the most significant operational challenge for VANETs is the requirement for improved MANET structure in order to keep up with the quick mobility of VANET nodes. This is the case because VANET nodes move more quickly than MANET nodes. This, in turn, needs better routing protocol design. Developing routing protocols that can handle vehicle ad hoc networks, which host a wide range of applications including security and welfare in intelligent transportation systems, has long been a challenge for researchers (ITS). However, the VANET has numerous qualities that distinguish it from the MANET, such as its high mobility, which is a feature that is unique to the VANET. There is no energy shortage for the cars in the VANET, but their mobility brings with it a host of new problems. Vehicles on the road are becoming increasingly scalable, making this issue more important. As a result of the enormous volume of messages sent between the vehicles involved in vehicle-to-vehicle communication, VANET suffers from significant packet loss. Some of the difficulties “VANET encounters include hidden terminals, long transmission delays for safety messages, broadcast storms, message security etc.”. A hierarchical organizational structure can be used to overcome these challenges. This occurs in hierarchical systems when two or more A. C. Pise (B) · K. J. Karande SKN Sinhgad College of Engineering, Korti, Pandharpur, India e-mail: [email protected] K. J. Karande e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 V. K. Gunjan et al. (eds.), Modern Approaches in Machine Learning and Cognitive Science: A Walkthrough, Studies in Computational Intelligence 1117, https://doi.org/10.1007/978-3-031-43009-1_25

305

306

A. C. Pise and K. J. Karande

cars with a shared characteristic create a cluster. Clustering is a common tool in data mining and machine learning. Mobile nodes in MANET, the forerunner to VANET, are similarly grouped for better performance. In a clustered vehicular environment, there are many smaller networks or clusters inside a larger network of cars.

2 Related Works Clustering was the most effective method that could arrange the functions of the system in a well-organized way. This strategy was used to create scalability in the network, minimize the amount of energy that was used, and produce a long lasting network lifespan. It is a well-known optimization issue that energy-efficient clustering has been generally recognized as playing a significant role in extending the lifespan of VANETs. This role has been extensively performed. The end purpose of clustering was to provide a clarification that maintains stability among the sensors in each and every network function.

2.1 Energy Efficient Clustering Methods Liu et al. [1] clustering and probabilistic broadcasting was the term given to the data transmission technique that was developed (CPB). The grouping was accomplished by clustering the automobiles according to the direction they were travelling in and their position. In this scenario, each vehicle makes its own judgment on the selection of CH in a distributed way; hence, the clustering management overhead is decreased as a result of this decision. Ahmad and Michel et al. [2], has put in place a clustering process known as Fan-Shaped Clustering (FSC). In terms of efficiency, this was a technique that used little power. VANETs favored this clusteringbased protocol over all others. This protocol used fan-shaped clusters to organize the whole network. All of these contributed to the network’s overall performance. Energy conservation and packet collection rates were shown to be superior in FSC to HEED in comparison to the HEED in the comparison results. Singh et al. [3], it has been demonstrated that clustering strategies for VANETs can be achieved through the use of a Genetic Algorithm-based approach. The clustering process was carried out taking into account both the amount of energy that was still present in the gateways and the distance that separated them from the sensor nodes. Singh et al. [4] Using type-2 fuzzy logic, we have developed a VANET algorithm. Interval type-2 FL model-based clustering is projected to perform better than Type-1 Fuzzy Logic (T1FL) model in cases of unclear decision levels. “The total number of levels in the sensor network was determined. Palvinder et al. [5] In VANETs, a clustering method based on LEACH (Low Energy Adaptive Clustering Hierarchy) has been explained. It was found that this strategy might increase the network’s stability

K-Mean Energy Efficient Optimal Cluster Based Routing Protocol …

307

and extend its lifespan through detailed simulations of the wireless sensors environment. In both dynamic and static networks, the “Orphan Nodes-Low Energy Adaptive Clustering Hierarchy (O-LEACH)” can be used. In order to reduce energy consumption, the LEACH approach was used. Vidhya and Sakthivel [6], Numerous use cases include the deployment of WSNs. “Once installed, sensor nodes might be difficult to charge because they are often located in difficult to access areas and WSN nodes are power constrained. An energy-saving routing approach is presented in conjunction with clusters in this article. Based on the LEACH algorithm, they provide an enhanced hierarchical routing chain-based clustering (EHRCC) technique. LEACH-1R PEGASIS (also known as E-LEACH) is a key component of the EHRCC algorithm”. Paola G. Vinueza et al. [7] have described a protocol for clustering in virtual private networks (VANETs). When it comes to multi-hop wireless sensor networks, the protocol known as the “Energy Aware Distributed Unequal Clustering” (EADUC) protocol was employed. This strategy was successful in resolving the entire energy problem. The residual energy and the location of the base station were given significant weight as clustering characteristics in this study. Srinivasa et al. [8], This study provides a new approach based on chemical reaction optimization (nCRO) for "energy-efficient CH selection and cluster formation."

2.2 Machine Learning Based Energy Efficient The machine learning played a significant role in society. Machine Learning algorithms can be used for minimizing energy consumption. It primarily helps the system to manage itself. VANETs have been implemented in several industrial applications, where consistency and network presentation were significant achievement factors. Some machine learning methods were as follows, Marappan et al. [9], For WSN data collecting, this study presents a “Cross LayerLow Energy Adaptive Clustering Hierarchy” (CL-LEACH) technique. This approach offers the energy usage of each network node, rather than the overall network, in order to enhance network longevity.

3 Proposed Energy Consumption Model and K-means Clustering Algorithms There is a transmitter and receiver in a VANET’s communication channel. The sensor node’s energy is used to power the amplifier circuit and amplifier circuit of the transmitter device. By radio electronics, the receiver can spend energy while receiving packets across long distances. During the communication period, the distance between the communication entities is compared to the threshold distance.

308

A. C. Pise and K. J. Karande

In this case, free space energy consumption is taken into account while performing the energy expenditure computation and the multipath design is employed otherwise. In light of this, the cluster head to sink node communication for data aggregation characteristics has been developed. The Eq. (1) can be used to calculate the amount of energy expended, which includes the distance d and the length L. eTX (L, D) = LeElE + LeFS D2 D ≪ DO

(1)

where, eFS can be used to describe the amount of money spent on transmitter amplifying expenses when the model is in free space, eTX This is the amount of energy expended to send one packet of data over a distance of L meters. The power consumption of the transmitter and receiver is mostly determined by the use of digital modulation and coding. eElE To operate single bits of data, an electrical circuit must expend energy in order to receive or transfer them. The power amplifier consumes electricity depending on the distance between the sender and receiver during communication (Wu et al., 2019). If the communication distance D is smaller than the D0 parameter, then the distance is considered to be short. The free space power loss model can then be evaluated when the energy calculation has been completed. Multipath mode can be considered for energy consumption if the distance is greater than the reference distance, as shown in the equation below, eTX (L, D) = LeElE + LeMP D4 D ≪ DO

(2)

where, eMP can be stated as the amount of money spent on transmitter amplifiers during multipath conditions. The following graph displays the energy usage from the receiving end, eRX (L, D) = LeElE

(3)

where, eRX up to the energy consumed to receive packet data at length L and L is defined as the number of bits in communication for data packet length. The formula for calculating D0 is as follows, √ D =√ 0

eFS

eMP

(4)

This section provides an overview of the energy consumption of the free space model and the multipath model. Sensor node operation must provide electricity for the transmitter and receiver. The K-means clustering approach helps reduce power consumption. The next section explains the K-means clustering method.

K-Mean Energy Efficient Optimal Cluster Based Routing Protocol …

309

4 Process of K-Means Clustering The primary goal of the proposed K-means energy efficient clustering is to reduce energy usage while still supporting scalability, allowing for the possibility of choosing a cluster head in close proximity, and reducing internal overheads. The selection of cluster nodes is a critical component in the clustering process because it ensures that cluster members may connect with the sink node in a fair manner without placing an undue burden on other nodes. The selection of the cluster head should be made via a run-time variable. At the end of each cycle, full nodes are updated to reflect the current state of the cluster head parameter in respect to the sink node. The K-means clustering method is used to produce optimally sized groups of clusters. The optimal packet size can be thought of as a decision variable for selecting the cluster head in a cluster of clusters of clusters. Knowing that longer packets can be introduced with greater loss ratings while small packets can be subjected to higher overload conditions is already a known fact. Furthermore, the fluctuation in packet size in networks of resource restrictions and autonomous constraints, which were not selected due to the increase in administration costs and additional overheads, has been observed. As a result, the appropriate packet size in this proposed methodology has been carefully considered in order to reduce the energy consumption of the WSN while simultaneously increasing the lifetime of the sensor nodes. The suggested K-means clustering algorithm can be implemented in three stages. Phase 1: Initialization The sensor nodes in the network field receive the broadcast message as an initialization request (IRQ) from the sink node at the start of the network field. The reply message to the sink node should be provided by the surviving node, which is referred to as the initialization reply message by the IETF (IRP). The IRP message also contains information about the nodes, such as the location of the sensors and the amount of energy consumed by the devices. Phase 2: Formation of Cluster In this phase, K-means clustering is used to formation of clusters; it can be simplest unsupervised learning methods. The clustering algorithm splitting dataset into K number of clusters in addition k value can be computed based on the Eq. (5). The K means clustering have more inter cluster in addition intra cluster similarity. The clusters have different process which are presented as follows, Step 1: Calculation of K Value Computation of the k value presented in the below equation, / K=

n 2π

/

εeFS F eMP XSN2

(5)

310

A. C. Pise and K. J. Karande

where, XSN2 can be “referred as the normal distance of all sensor nodes to the sink node, F can be described as the dimension of the created network structure and n can be referred as the sensor nodes in the network structure. The k value only selected the number of clusters in the network structure”. Step 2: Computation of Distance Here, “Euclidean distance can be utilized to compute the distance among the sensor nodes to each of the cluster centers. The Euclidean distance should assign each point to the closest center. The Euclidean distance presented below equation”,

XCC

[ | n |Σ = | (Xa − Xc )

(6)

a=1

where, Xc can be described as the coordinates of cluster centers, Xa can be described as the coordinate of sensor node a in addition XCC can be described as the distance of node to cluster center. Step 3: Computation of New Cluster Center The calculation of the mean values in each cluster results in the creation of a new cluster center in each of the cluster groups’ sensor nodes that are completely connected. Step 4: Termination Condition Following the computation of the new centers in the different clusters, the step 2 can be carried out using the newly computed centers in each cluster. If the sensor nodes for cluster formation vary, then step 3 should be completed; otherwise, the method should be terminated. Phase 3: Selection of Cluster Head When using k means clustering, the selection of the cluster head is a critical step in the cluster building process. Clusters are generated as a result of the two phases that have gone before. Following that, a cluster head might be selected from among the clusters during this phase. The cluster head can be chosen depending on the two separate weight functions that have been defined, Wna = S1 ∗ Ena + S2 ∗ DCC

(7)

WS = S1 ∗ eTX + S2 ∗ Avg(DCC )

(8)

where, eTX can be referred as the energy used up to transmitting packet data at the length L, WS can be described as the cluster head with the standard weight of a node, DCC can be described as the distance among the ath the node to the cluster center and S1 , S2 are considered as the constants the a = 1, 2, 3 . . . n. .

K-Mean Energy Efficient Optimal Cluster Based Routing Protocol …

311

5 Evaluation Metrics Throughput: The accomplishment of acknowledged packets via a communication channel can be characterized as the throughput of that communication channel (Fig. 1 and Table 1). Packet to the Delivery ratio (PDR): When compared to the total number of packets that were sent across all channels of communication, the number of packets that were delivered properly and correctly is much higher (Fig. 2). Energy Consumption: In the field of computing, this term refers to the entire amount of energy that is used up throughout the process of moving packets from their origin to their final destination (Fig. 3).

Fig. 1 Throughput vehicles is 200

Number of Vehicles is 200 8200 8000 7800 7600 7400 7200 7000 6800 6600 6400

Throughput (kbps)

Table 1 Performance metrics results

Existing system

No of vehicle

Throughput (kbps)

Packet to the delivery ratio (%)

Energy consumption (J)

50

8524

59.55

252

10,547

92.75

189

7284

61.75

242

8247

89.64

156

7025

56.72

342

8115

82.78

220

Proposed system Existing system

100

Proposed system Existing system Proposed system

200

312 Fig. 2 PDR vehicles is 200

A. C. Pise and K. J. Karande

Number of Vehicles is 200 100 80 60 40 20

Packet to the Delivery ratio (%)

0

Fig. 3 Energy consumption (J) vehicles is 200

Number of Vehicles is 200 400 350 300 250 200 150

Energy Consum ption (J)

100 50 0

6 Conclusion This is the MANET area that is currently in the process of being developed. IT’s is one of the most demanding applications on the VANET, and it is widely regarded as the most effective method of implementing IT’s on the VANET. Because of the reduced travel time, VANET encourages people to be healthier, more tranquil, and more joyful in their lives. We have discussed VANETs in basic terms, including their uses and characteristics, in this review paper, and we have addressed the important research concerns that must be considered in the creation of cost-effective VANET protocols and applications (IT’s). Other difficulties must be considered, but the general view

K-Mean Energy Efficient Optimal Cluster Based Routing Protocol …

313

is that vehicles will soon be able to benefit from spontaneous wireless connections, which will enable VANETs to become a reality. In this paper we study various conventional and machine learning algorithms for VANET’s also we analysis the K-mean clustering algorithms for energy efficient routing and cluster head selection.

References 1. Liu, L., Chen, C., Qiu, T., Zhang, M., Li, S., & Zhou, B. (2018). A data dissemination scheme based on clustering and probabilistic broadcasting in VANETs. Vehicular Communications. 2. Abuashour, A., & Kadoch, M. (2017). Performance improvement of cluster-based routing protocol in VANET. IEEE Access, 5, 15354–15371. 3. Mann, P. S., & Singh, S. (2017). Artificial bee colony metaheuristic for energy-efficient clustering and routing in VANETs. An International Journal of Soft Computing, 21(22), 6699–6712. 4. Mann, P. S., & Singh, S. (2017). Energy efficient clustering protocol based on improved metaheuristic in VANETs. An International Journal of Network and Computer Applications, 83, 40–52. 5. Mann, P. S., & Singh, S. (2017). Energy-efficient hierarchical routing for wireless sensor networks: a swarm intelligence approach. Wireless Personal Communications, 92(2), 785–805. 6. Vidhya, G., & Sakthivel, S. (2017). Energy-efficient enhanced hierarchical routing chain based clustering for wireless sensor networks. Wireless Personal Communications, 92(2), 785–805 (2017) 7. Naranjo, P. G. V., Shojafar, M., Mostafaei, H., Pooranian, Z., & Baccarelli, E. (2017). P-SEP: A prolong stable election routing algorithm for energy-limited heterogeneous fog-supported VANETs. An International Journal of Supercomputing, 73(2), 733–755. 8. Srinivasa, P. C., Rao, & Banka, H. (2017). Energy efficient clustering algorithms for VANETs: Novel chemical reaction optimization approach. An International Journal of Wireless Networks, 23(2), 433–452. 9. Preetha, M., & Rodrigues, P. (2016). An energy efficient routing protocol for correlated data using CL-LEACH in WSN. An International Journal of Wireless Networks, 22(4), 1415–1423.

Dandelion Algorithm for Optimal Location and Sizing of Battery Energy Storage Systems in Electrical Distribution Networks Rajesh Patil and Varaprasad Janamala

1 Introduction In recent times, the problem of energy storage system (ESS) allocation in electrical distribution networks (EDNs) has received high attention for achieving energy equilibrium, autonomous microgrid (MG) operation, handling various operational and controlling issues, and reliability improvement [1]. In [2], superconducting magnetic energy storage (SMES) is combined with model predictive control and the PV system to provide reactive power support and improve the stability of the utility’s voltage. In [3, 4], a bi-level optimization approach is suggested for sizing the ESS, including its reactive power support, to deal with high RE uncertainty and penetration. In the outer optimization, life cycle cost (LCC) is taken into account, and in the inner optimization, voltage profile and loss are taken into account. In [5], hybrid RE systems with PV and WT are set up so that they work best both on and off the grid. Loss of load and emission costs are optimized by using spotted hyena optimization (SHO) and particle swarm optimization (PSO) while designing the ESS. In [6], voltage fluctuation and capacity of ESS are optimized in the distribution system by using multi-objective inertia control PSO for moderating the power exchanges between variable PV power and loading conditions. In [7], the uncertainty of PV, WT, and load is mitigated via the optimal design of the ESS, considering reduction of energy loss and enhancement of voltage stability using the rider optimization algorithm (ROA). In [8], the problem of ESS allocation is solved using MINP and optimized using PSO. The cost

R. Patil (B) SKN Sinhgad College of Engineering, Pamdharpur, Maharashtra, India e-mail: [email protected] V. Janamala CHRIST (Deemed to be University), Bangalore, Karnataka, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 V. K. Gunjan et al. (eds.), Modern Approaches in Machine Learning and Cognitive Science: A Walkthrough, Studies in Computational Intelligence 1117, https://doi.org/10.1007/978-3-031-43009-1_26

315

316

R. Patil and V. Janamala

of energy not supplied (ENS) and the total cost of ESS are considered when formulating a multi-objective function. In [9, 10], teaching–learning-based optimization (TLBO) is used to determine the best number of locations and their sizes, as well as the financial elements of ESS (including energy not delivered, investment cost, and operating cost) and their impact on active power distribution losses and dependability. In [11], a multi-objective ESS allocation problem by the non-dominated sorting genetic algorithm II (NSGA-II) is solved to maximize reliability, specifically the momentary average interruption frequency index (MAIFI) and system average interruption frequency index (SAIDI), as well as equipment cost. For managing uncertainties with wind turbines (WT) and fuel cells (FC) in microgrid operations while taking the risk of downstream failures into consideration, an improved grey wolf optimization (GWO) is presented in [8]. In [12], modified manta ray foraging optimizer (MMRFO) is employed for sizing the ESS for different combinations of PV and WTs in the EDNs, considering their uncertainties. In [10], coyote optimization algorithm (COA) is used for solving BES in an islanding network considering EV load penetration. Further, a comprehensive review on different state-of-the-art optimization studies on ESS allocation can be found in [13]. In light of the above works, this paper introduces a new optimization approach based on Dandelion Algorithm (DA) [14] for optimizing the performance of EDN via optimally integrating distributed battery energy storage systems (BESSs). At first, the search space for BESs locations is predetermined using loss sensitivity factors (LSFs) and later, DA is utilized for deducing the optimal locations along with sizes. Real power distribution loss reduction is considered as major objective function and the impact of BESs is further extended to analyze the network voltage profile, voltage stability and GHG emission. For calculating the computational effectiveness of the suggested methodology, IEEE 33-bus RDN is used. The findings produced demonstrate DA’s efficacy in comparison to other heuristic algorithms. Additionally, the results show that the proposed DA significantly improves all techno-economic-environmental factors and improves RDN performance.

2 Battery Energy Storage System The basic objective of BES system in this research is to handle sporadic nature of RE generation and the random network loading profile. In the event of excessive generation, the BES system can store that excessive generation in charging mode, or at times of deficit power generation by the RE source, it acts like another dispatchable power source by discharging its storage energy. Thus, the BES system capacity needs to be optimized considering autonomous or islanding modes. Following mathematical model is employed for designing BES capacity [10]. B E S cp =

L F × PD,max × t αt × βa × D0 D × ηin × Vbs

(1)

Dandelion Algorithm for Optimal Location and Sizing of Battery …

317

where BES cp is the battery energy storage capacity in Ah, α t is the correction factor for temperature changes, β a is the correction factor for ageing, DoD is the depth of discharge allowed, V bs is the battery specified voltage, PD,max is the maximum load of the network, t is the time duration for back-up, ηin is the battery inverters’ efficiency.

3 Problem Formulation Loss minimization is considered the major objective function in this research, which involves maintaining bus voltage magnitude limits and branch currents within their specified limits.  F = min Ploss =

nbr 

 2 Ibr rbr

(2)

Ibr = Ibr (max) , br = 1 : nbr

(3)

Vn(min) ≤ Vn ≤ Vn(max) , n = 1 : nbus

(4)

B E S cp(min) ≤ B E S cp,k ≤ B E S cp(max) , k = 1 : nbes

(5)

br =1

Subjected to:

where Ploss is the real power loss of the network, I br is the branch-k current, V n is the bus voltage magnitude, I br(max) is the maximum current limit, V n(min) and V n(max) are the low and high voltage limits, respectively; BES cp,k, BES cp(min) and BES cp(max) are the BES capacity at bus-k, and its low and high limits, respectively; nbr, nbus, and nbes are the number of branches, number of buses and BES locations in the network, respectively.

4 Dandelion Algorithm The ground is divided into two groups, according to the dandelion algorithm (DA) [15]: those that are best for dandelion sowing and those that are not. The core dandelion (CD) is the dandelion that grows in an advantageous environment, and assistance dandelions are those that do not (AD). When a dandelion is sown, its seeds are scattered all over the surrounding area. The dandelion-sowing process in DA can be shown to look for an optimal in a particular region close to a point. Besides CD and AD, different dandelions sow differently. Mutant sowing prevents settling into the

318

R. Patil and V. Janamala

local optimum. The selection mechanism chooses dandelions for the next generation. The dandelion algorithm includes normal, mutant sowing, and selection strategies.

4.1 Normal Sowing The CD produces more seeds than the AD because its soil is more conducive to seed development. The sowing yield depends on the fitness of the dandelion population. A low-fitness dandelion sows more seeds, whereas a high-fitness one sows fewer seeds. However, neither dandelion will sow fewer seeds than the least possible number of seeds, as determined by,  ns i =

− f (di )+ε ns max × ffmax ns i > ns min max − f min +ε ns max ns i ≤ ns min

(6)

where nsmax and nsmin are the maximum and minimum number of seeds, respectively; f max = max((d i )) and f min = min((d i )), ε is machine epsilon to avoid a 0 denominator, ns i is the number of seeds, di represents ith dandelion. As mentioned earlier, dandelions are separated into two types as AD and CD. The CD is the highest fitness dandelion, determined by subtracting the fitness of the ADs from the fitness of the CD. Dcd = min( f (di ))

(7)

The radius of the ADs and the central dandelions are calculated differently. The sowing radius of the AD (except for CD) is determined by:  rad(k) =

u b − lb k = 1 ω × rad(k−1) + (Dcd  − di ∞ ) else

(8)

where k is the number of current iteration, ub and l b are the upper and lower limits of objective function, respectively. From (8), the ADs’ seeding radius is set to the search space’s diameter. Then, it’s set to the distance between the current AD and the CD. DO include the planting radius of AD in the last iteration, where ω is a weighting factor, for slowing down convergence and boost global search efficiency, ω is designed as follows.

4.2 Mutation Sowing for the Core Dandelion CD suggests mutation planting to stop local optimum and keep population diversity, as explained by,

Dandelion Algorithm for Optimal Location and Sizing of Battery …

Dcd = Dcd × (1 + Levy())

319

(9)

where Levy() is a Levy distribution random number with parameter 1.5. DO necessitates keeping the best place for the future generation. The remaining locations are chosen using a disruptive selection operator to maintain variation. For di , probability pi is calculated by,    f i − f avg  pi =  S D n=1 f n

(10)

where SD is the number of all dandelions (normal seeds and mutant seeds), f i is the objective function’s fitness value, and f avg is the generational mean fitness value. People with average fitness scores will be thrown out, but both good and bad people may have a better chance of being picked in the next round. This method preserves population diversity while improving global search.

5 Results The performance of DA is compared with that of AOA, FSA, PFA, and BOA. The convergence criterion for all algorithms is the maximum number of iterations.

5.1 Simulations on IEEE 33-Bus Test System The base case test system has 33 nodes. The lowest voltage magnitude is registered at 0.9039 p.u. at bus 18. The voltage stability index (VSI) was observed at 0.6675. Also, the GHG emission and AVDI are estimated at 8.0382 e6 lb/hr and 0.0545 p.u., respectively. Using the proposed DA, the best way to connect three BESSs is planned by only thinking about how to minimize real power loss. At first, the LSIs are determined to identify the top ten most preferable locations. Later, DA is implemented for determining optimal locations and sizes. The size of each BESS is limited to 2000 kW. In the same way as with DA, Table 1 shows the best results from 50 independent runs of different algorithms. According to F value, BOA, PFA, and DA are very competitively produced at the global optimum. DA, on the other hand, performed well when the best, worst, median, and standard deviation were evaluated. As per computation time, DA has shown better computational features. From this, it can be

320

R. Patil and V. Janamala

Table 1 Solution by different algorithms in IEEE 33-bus Method

Sizes in kW (bus #)

Objective function Best

Worst

Median

Std.

Time (s)

1082.81 (24) 774.95 (14) 1077.37 (30)

71.815

131.580

71.818

11.221

2.1325

FSA

1077.09 (24) 1065.76 (30) 804.70 (13)

71.838

121.908

72.252

9.718

2.8437

PFA

1082.66 (24) 774.94 (14) 1077.70 (30)

71.815

103.427

72.029

7.232

2.6526

AOA

1065.96 (30) 1077.41 (24) 804.68 (13)

71.838

114.609

71.895

7.015

2.1125

Proposed

1071.01 (30) 1099.92 (24) 754.49 (14)

71.457

103.945

72.858

6.326

2.0158

Fig. 1 Convergence characteristics in IEEE 33-bus RDS

Loss (kW)

BOA

139.500 129.500 119.500 109.500 99.500 89.500 79.500 69.500

BOA FSA PFA AOA DA

1 5 9 13 17 21 25 29 33 37 41 45 49 Iteration

seen that DA is resulting in global optima. The convergence characteristics are shown in Fig. 1. In addition, the comparison of the voltage profile is depicted in Fig. 2. Equation (1) tells us how much power a battery can hold if we assume that LF is 0.25 over a year. From [10], the battery parameters are taken as follows: t = 0.964, a = 0.85, DoD = 0.8, in = 0.95, and V bs = 48 V. With these parameters, the required BESS capacity is equal to 272.18 MWh.

5.2 Simulations on IEEE 69-Bus Test System The base case test system has 69 nodes. The lowest voltage magnitude is registered at 0.9092 p.u. at bus-65. The voltage stability index (VSI) is observed at 0.55. Also,

Dandelion Algorithm for Optimal Location and Sizing of Battery …

Voltage Magnitude (p.u.)

Fig. 2 Voltage profile of IEEE 33-bus RDS

1.020 1.000 0.980 0.960 0.940 0.920 0.900 0.880 0.860

321

Base

BESS

1 4 7 10 13 16 19 22 25 28 31

Fig. 3 Convergence characteristics in IEEE 69-bus RDS

Loss (kW)

Bus Number

130.000 120.000 110.000 100.000 90.000 80.000 70.000 60.000

BOA AOA PFA FSA DA

1 5 9 13 17 21 25 29 33 37 41 45 49 Iteration

the GHG emission and AVDI are estimated at 8.2461 e6 lb/hr and 0.0014 p.u., respectively. Three BESSs are optimally planned to integrate using the proposed DA, considering only real power loss minimization. At first, the LSIs are determined to identify the top ten most preferable locations. Later, DA is implemented for determining optimal locations and sizes. The size of each BESS is limited to 2000 kW. In a similar manner to DA, other algorithms, namely AOA, FSA, PFA, and BOA, are also simulated. The best results obtained over 50 independent runs are given in Table 1. According to objective function value, BPA, PFA, and DA are very competitively produced at the global optima. DA, on the other hand, performed well when best, worst, median, and standard deviation were evaluated. As per computation time, DA has shown better computational features. From this, it can be seen that DA is resulting in global optima. The convergence characteristics of these algorithms and the comparison of voltage profiles are given in Figs. 3 and 4, respectively. In a similar manner to IEEE 33-bus, the BESS parameters are taken as the same for this test system also. The required BESS is estimated at 278.56 MWh (Table 2).

322

R. Patil and V. Janamala

Fig. 4 Voltage profile of IEEE 69-bus RDS Voltage (p.u.)

1.050 1.000 0.950 Base BESS

0.900 0.850

1 6 1116 212631 364146 51566166 Bus Number

Table 2 Solution by different algorithms in IEEE 69-bus Method

Sizes in kW (bus #)

Objective function Best

Worst

Median

Std

Time (s)

BOA

370.21 (69) 1728.31 (61) 492.39 (66)

73.066

102.938

73.073

5.998

2.356

FSA

1703.79 (61) 564.29 (51) 466.17 (18)

70.178

117.765

70.178

9.384

2.935

PFA

328.07 (69) 398.74 (17) 1749.41 (61)

70.144

106.317

70.335

5.914

2.278

AOA

826.36 (69) 791.60 (17) 1733.84 (61)

72.999

110.063

72.999

6.082

2.457

Proposed

1719.45 (61) 525.74 (11) 380.97 (18)

69.428

90.274

69.436

5.264

2.156

6 Conclusion This paper presents a new optimization method based on the Dandelion Algorithm (DA) for improving EDN performance by integrating distributed battery energy storage systems (BESs) in the best way possible. The main goal is to cut down on real power distribution loss, and the effects of BES are also looked at in terms of the network voltage profile, voltage stability, and GHG emissions. RDNs are used to figure out how well the proposed method works in terms of computation. Based on the results, it’s clear that DA works better than other heuristic algorithms. Also, the proposed DA improves all technological, economic, and environmental factors as well as the performance of the RDNs.

Dandelion Algorithm for Optimal Location and Sizing of Battery …

323

References 1. Zidar, M., Georgilakis, P. S., Hatziargyriou, N. D., Capuder, T., & Škrlec, D. (2016). Review of energy storage allocation in power distribution networks: Applications, methods and future research. IET Generation, Transmission and Distribution., 10(3), 645–652. 2. Bakeer, A., Salama, H. S., & Vokony, I. (2021). Integration of PV system with SMES based on model predictive control for utility grid reliability improvement. Protection and Control of Modern Power Systems., 6(1), 1–3. 3. Emara, D., Ezzat, M., Abdelaziz, A. Y., Mahmoud, K., Lehtonen, M., & Darwish, M. M. (2021). Novel control strategy for enhancing microgrid operation connected to photovoltaic generation and energy storage systems. Electronics, 25;10(11), 1261. 4. Arabi-Nowdeh, S., Nasri, S., Saftjani, P. B., Naderipour, A., Abdul-Malek, Z., Kamyab, H., & Jafar-Nowdeh, A. (2021). Multi-criteria optimal design of hybrid clean energy system with battery storage considering off-and on-grid application. Journal of Cleaner Production., 25(290), 125808. 5. Khasanov, M., Kamel, S., Rahmann, C., Hasanien, H. M., & Al-Durra, A. (2021). Optimal distributed generation and battery energy storage units integration in distribution systems considering power generation uncertainty. IET Generation, Transmission and Distribution, 15(24), 3400–3422. 6. Saboori, H., Hemmati, R., & Jirdehi, M. A. (2015). Reliability improvement in radial electrical distribution network by optimal planning of energy storage systems. Energy, 15(93), 2299– 2312. 7. Lata, P., & Vadhera, S. (2020). Reliability improvement of radial distribution system by optimal placement and sizing of energy storage system using TLBO. Journal of Energy Storage, 1(30), 101492. 8. Miao, D., & Hossain, S. (2020, Jul 1). Improved gray wolf optimization algorithm for solving placement and sizing of electrical energy storage system in micro-grids. ISA Transactions, 1(102), 376–387. 9. Abdel-Mawgoud, H., Ali, A., Kamel, S., Rahmann, C., & Abdel-Moamen, M. A. (2021, Jun 24). A modified manta ray foraging optimizer for planning inverter-based photovoltaic with battery energy storage system and wind turbine in distribution networks. IEEE Access, 24(9), 91062–91079. 10. Janamala, V., & Reddy, D. S. (2021, Sep 1). Coyote optimization algorithm for optimal allocation of interline–Photovoltaic battery storage system in islanded electrical distribution network considering EV load penetration. Journal of Energy Storage, 1(41), 102981. 11. Pombo, A. V., Murta-Pina, J., & Pires, V. F. (2017, Jul 1). Multiobjective formulation of the integration of storage systems within distribution networks for improving reliability. Electric Power Systems Research, 148, 87–96. 12. Abdel-Mawgoud, H., Ali, A., Kamel, S., Rahmann, C., & Abdel-Moamen, M. A. (2021, Jun 24). A modified manta ray foraging optimizer for planning inverter-based photovoltaic with battery energy storage system and wind turbine in distribution networks. IEEE Access, 9, 91062–79. 13. Giridhar, M. S., Radha Rani, K., Sobha Rani, P., & Janamala, V. (2022). Mayfly algorithm for optimal integration of hybrid photovoltaic/battery energy storage/D-STATCOM system for islanding operation. International Journal of Intelligent Systems, 15, 225–232. 14. Yang, B., Wang, J., Chen, Y., Li, D., Zeng, C., Chen, Y., Guo, Z., Shu, H., Zhang, X., Yu, T., & Sun, L. (2020, Dec 1). Optimal sizing and placement of energy storage system in power grids: A state-of-the-art one-stop handbook. Journal of Energy Storage, 32, 101814. 15. Zhao, S., Zhang, T., Ma, S., & Chen, M. (2022, Sep 1). Dandelion optimizer: A nature-inspired metaheuristic algorithm for engineering applications. Engineering Applications of Artificial Intelligence, 114, 105075.

A Survey of Internet of Things Frameworks for Crowd Management System Jyoti Ambadas Kendule and Kailash Karande

1 Introduction Crowd management refers to the choices and actions made to monitor and regulate heavily populated areas, and it entails a number of issues, from situational awareness and evaluation to the implementation of activities that are appropriate for the time being. Crowd control has grown more challenging, particularly when social distance is required. These problems can be solved with the use of Internet of Things (IoT) technologies. Crowd management is the well-ordered organization and control of the lawful gathering and movement of people, and it entails the evaluation of individuals who test the viability of areas before their usage. Crowd management is the deliberate and systematic provision of direction for the orderly movement of huge crowds. As part of crowd management, steps can be done to limit or control how people behave in large groups. Crowd management is the term for this, which may involve securing public safety. Additionally, crowd management may be described as a series of preparations and actions made to facilitate, utilize, and move crowds. The structure of this survey article is as follows: In part 1, we discussed the basic concepts related to IoT. We detailed some great and often technologies related to IoT in part 2. The parameters like protocols, applications of IoT were outlined in part 3 of this series. Part 4 contains the some IoT based framework for crowd management system. Part 5 contains the overall conclusion.

J. A. Kendule (B) Electronics and Telecommunication, SVERIs College of Engineering, Pandharpur, India e-mail: [email protected] K. Karande Electronics and Telecommunication, SKN Sinhgad College of Engineering, Pandharpur, India © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 V. K. Gunjan et al. (eds.), Modern Approaches in Machine Learning and Cognitive Science: A Walkthrough, Studies in Computational Intelligence 1117, https://doi.org/10.1007/978-3-031-43009-1_27

325

326

J. A. Kendule and K. Karande

Fig. 1 Introduction to IoT

2 Basics of IoT The phrase “Internet of Things” (IoT) has been in use for quite a few years. It is gaining ground in this environment as improved wireless technology evolves. The presence of various things, including RFID, NFC, sensors, actuators, and mobile phones, is the fundamental tenet of this notion. IoT is used to incorporate a variety of market-available technologies, including RFID, machine-to-machine communication, and vehicle-to-vehicle communication. The term “Internet of Things” describes the coding and networking of commonplace items and things to make each one machine-readable and traceable on the Internet (Fig. 1). IoT functions through a concoction of physical devices, cloud computing, advanced data analytics, and wireless networking technology. IoT functions in a basic manner as shown in the following: • A collection of physical objects is wired or wirelessly connected to one another or to a central location. • The gadgets use some sort of sensor to get information from the outside environment. • Then, the data is kept somewhere, be it on the device, in the cloud, or at a middletier network site. • Following that, the data is processed, frequently using artificial intelligence and machine learning. • The physical gadget employs the processed data to execute a specific action.

3 Functional Blocks A multitude of functional building blocks make up an IoT system, giving it the capacity to recognize, perceive, act, communicate, and manage. These are the structural units:

A Survey of Internet of Things Frameworks for Crowd Management …

327

• Device: Perform sensing, actuation, monitoring, and control tasks make up an IoT system. • Communication: Manages communication in the IoT system. • Services: This block gives services like device control, device monitoring, data transmission, and device discovery. • Management: Several controls for the IoT system are provided by this block. • Security: This block safeguards the IoT system by offering different functions related to data security. • Application: Users can use this interface to manage and keep an eye on various parts of the IoT system. Users of the application can also see the position of the system and examine the processed data.

4 Protocols Within a certain IoT protocol, the application layer acts as the interface between the user and the device. The transport layer in every IoT protocol facilitates and secures data transmission as it passes between layers. An IoT protocol’s network layer facilitates device-to-router communication. The IoT protocol’s data layer is responsible for transferring data within the system architecture as well as detecting and fixing any errors that were discovered in the physical layer. The physical layer serves as the communication pathway for gadgets inside a given setting.

5 Applications The Internet of Things is being employed in many other industries, such as agriculture, smart homes, healthcare, and the environment (Fig. 2).

6 IoT Based Framework for Crowd Management System One of the major objective and contribution of this paper is to present a survey for IoT based framework for crowd management system. In this section some papers are examined on the basis of framework design for crowd management system. Numerous researchers have presented their work on crowd control and helped to advance the field of crowd control (Fig. 3). a. Mohammad Yamin et al. presented a paper [1]. This approach is for managing crowded events during the current epidemic might also be applied to other crises and pandemics that are brought on by either natural disasters or human errors. This framework makes use of information on COVID-19 transmission prevention

328

Fig. 2 IoT Applications

Fig.3 IoT framework—smart city

J. A. Kendule and K. Karande

A Survey of Internet of Things Frameworks for Crowd Management …

b.

c.

d.

e.

f.

g.

329

found on various World Health Organization web portals. This study offers a detailed algorithm as well as measurements for verifying some of the framework’s recommended components. Waskitho Wibisono et al. presented a research [2]. This paper addresses the design and implementation of a mobile crowdsensing framework. The suggested framework’s prototype has been successfully used and tested with actual framework, and the outcomes have been encouraging. Jose Joaquin Peralta Abadia et al. presented survey [3]. This paper compares and summarizes the IoT frameworks technology for smart city applications. It also presents review on these frameworks. Imran Ahmed et.al presented an article [4]. Article introduces a crowd surveillance system built on the Internet of Things that employs a deep learning model to identify and count people from an overhead perspective. For the purpose of detecting persons, the Single Shot Multibox Detector (SSD) model with Mobilenetv2 as the fundamental network is utilized. Without and after additional training, the model’s counting accuracy is 92% and 95%, respectively. The created system may be useful for several applications related to crowd management. Marwa F. Mohamed et al. presented a research [5]. This article introduces a crowd management strategy for spreading visitors throughout a busy space. The sensors, management, and interaction layers make up the framework. The suggested system will efficiently reduce time and assist administrators distribute and control visitors using inexpensive sensors controlled by a smartphone. Mohammad Yamin et al. presented an article [6]. This article outlines the design of a health management system and crowd control that was created specifically to avoid and control crowd disasters. The system is divided into two parts, one is for managing disasters and the other for managing healthcare. A suggested system includes an algorithm for the early detection of disasters. Soumya Kanti Datta et al. presented this paper [7]. An effective MCS-enabled IoT framework is presented in this study for smart cities. Cooperative crowd sensing and a data-centric strategy can offer a unified mechanism to address numerous issues that smart cities face. Additionally, talk about incorporating the aforementioned IoT framework into a single M2M standard architecture. The work’s primary contribution is the creation of a comprehensive framework for the development of context-aware, power-aware mobile applications. The IoT framework, its component systems, and how they are all integrated into a single M2M architecture are then covered.

Some popular Internet of Things (IoT) frameworks that can be used for crowd management systems: 1. Microsoft Azure IoT Suite: Azure IoT Suite is a comprehensive cloud-based platform that provides various IoT services and features. It offers real-time analytics, device management, and integration with other Azure services. It provides robust security features and supports a wide range of devices and protocols.

330

J. A. Kendule and K. Karande

2. IBM Watson IoT Platform: IBM Watson IoT Platform is a fully managed, cloud-based IoT solution. It offers device connectivity, data storage, analytics, and visualization tools. It provides advanced analytics capabilities and supports integration with other IBM Watson services, such as AI and machine learning. 3. Google Cloud IoT Core: Google Cloud IoT Core is a scalable and secure IoT platform provided by Google Cloud. It allows you to connect, manage, and ingest data from IoT devices. It provides real-time analytics, integration with other Google Cloud services, and supports industry-standard protocols. 4. AWS IoT Core: AWS IoT Core is a managed cloud platform offered by Amazon Web Services (AWS) for building IoT applications. It enables device connectivity, secure data communication, and device management. It integrates with other AWS services, such as AWS Lambda for serverless computing and AWS IoT Analytics for data analysis. 5. ThingWorx: ThingWorx is an IoT platform developed by PTC. It provides features like device connectivity, data collection, analytics, and visualization tools. ThingWorx supports rapid application development using drag-and-drop components and offers integration with other enterprise systems. 6. Kaa IoT Platform: Kaa is an open-source IoT platform that offers device management, data collection, and analytics capabilities. It supports various IoT protocols and provides customizable components to build IoT applications. Kaa also provides security features and can be deployed on-premises or in the cloud. 7. Bosch IoT Suite: Bosch IoT Suite is an open-source-based IoT platform developed by Bosch. It provides features like device management, connectivity, data processing, and analytics. It offers integration with other Bosch products and services, making it suitable for industrial IoT applications.

7 Conclusion IoT is now being used everywhere that affects people, including smart cities, healthcare security and emergencies, smart environments, smart business processes, smart agriculture, domestic and home automation. We discussed the technologies and their specifications in this article in order to make the Internet of Things a reality. Following that, we give some excellent examples of how the Internet of Things may be quite useful. This study offers an overall analysis of crowd studies, covering everything like crowd detection, crowd monitoring and crowd management. The safety of the crowd is the main objective of crowd-related research.

References 1. Almutairi, M. M., Yamin, M., Halikias, G., & Abi Sen, A. A. (2022). A framework for crowd management during COVID-19 with artificial intelligence: Sustainability, 14(1).

A Survey of Internet of Things Frameworks for Crowd Management …

331

2. Wibisono, W. (2017). Proceedings of the 14th EAI international conference on mobile and ubiquitous systems”, computing, networking and services. 3. Joaquín, J. (2022). A systematic survey of internet of things frameworks for smart city applications. Sustainable Cities and Society. 4. Ahmed, I., Ahmad, M., Ahmad, A., & Jeon, G. (2021).: IoT-based crowd monitoring system: Using SSD with transfer learning. Computers and Electrical Engineering. 5. Mohamed, M. F., Shabayek, A. E. R., & El-Gayyar, M. (2019). IoT-based framework for crowd management: Springer innovations in communication and computing. 6. Yamin, M., Basahel, A. M., & Abi Sen, A. A. (2018). Managing crowds with wireless and mobile technologies: Hindawi. Wireless Communications and Mobile Computing, 2018, Article ID 7361597. 7. Datta, S. K., Da Costa, R. P. F., Bonnet, C., & Härri, J. (2016). One M2M architecture based IoT framework for mobile crowd sensing in smart cities.

Performance Analysis of Patient Centric EHR Through Hyperledger Fabric Somnath Agatrao Zambare and Namdev M. Sawant

1 Introduction Electronic Medical Records (EMRs) are patient records that can be created, collected, managed and accessed by licensed medical and nursing personnel treatment. Providing a standard of security and personal access is one of the most important aspects of healthcare. In the age of big data, we use big data to store and access a lot of health information on the Internet. Cloud computing has played an important role in this process. Electronic health record (EHR) systems are easy to use and reliable, but raise many privacy and security concerns contains confidential information. It is the most sensitive data collection method. The development of the Internet and digital health systems has made electronic health records vulnerable to security breaches. Therefore, the security and confidentiality of EHR data should be considered when evaluating classification based on reliability [1, 2].

1.1 Types of Health Record I. PHR: And electronics application software that permits to keep and control their own health record in private environment with secure and confidential manner known as Personal Health Record (PHR). S. A. Zambare (B) · N. M. Sawant SKNSCOE, Pandharpur, Maharashtra, India e-mail: [email protected] N. M. Sawant e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 V. K. Gunjan et al. (eds.), Modern Approaches in Machine Learning and Cognitive Science: A Walkthrough, Studies in Computational Intelligence 1117, https://doi.org/10.1007/978-3-031-43009-1_28

333

334

S. A. Zambare and N. M. Sawant

II. EMR: An electronic medical record (EMR) is a digital copy of all information normally found in a health care facility’s paper records: medical history, diagnoses, medications, immunization records, allergies, results Tests and notes of a doctor in a medical facility. III. EHR: The digital form of patient’s health record is known as an Electronic Health Record (EHR). It is also called as digital version of a paper based patient record. The EHR is a patient-centric, real-time recording that makes information immediately and safely available to authorized users.

1.2 Problems with Current EMR Systems The healthcare hospitals, institutions and industries related to healthcare sectors are prime target for attackers due to the lack of technical expertise in the industry. Recent strikes on the healthcare industry highlight data safety challenges in the sector. Targeted strikes include phishing attacks and ransomware that obtain the information successfully but it is not limited to get personal data. In fact, the high success rate of ransomware strikes indicates a lack of basic security measures such as backups and updation in system. Medical applications such as EMR are very tactful because they directly involve individual and sensitive data that must be protected from unofficial access [3].

2 Blockchain Technology 2.1 What is Blockchain? Blockchain is a system that records information in such a way that it is difficult or impossible to change, control or disable the system. A blockchain is essentially a digital record of transactions distributed over a network of computer systems. Each block in the chain contains the cryptographic hash of the previous block, and when a new transaction occurs on the blockchain, information about the transaction is added to each participant’s ledger. A block with a name and content [4] (Fig. 1). The Block Header Contains: • Version: Specifies the validation rules for the block. • Hash of Previous block: This allows the old block to remain unchanged. • Timestamp: The current time the block was created.

Performance Analysis of Patient Centric EHR Through Hyperledger Fabric

335

Fig. 1 Structure of blockchain

• Merkle root hash: The Merkle root is obtained from the hash value of every transaction in the block. • Nonce: Nonce is a unique 4-byte number used only once in a message. This will be added to the karma that meets the difficulty limit.

2.2 Types of Blockchain 2.2.1

Public Blockchain

Public blockchains are permission less in nature, open to everyone, and fully decentralized. It allows equal rights to all nodes present in the blockchain to access the blockchain & to create new data blocks, verify data blocks.

2.2.2

Private Blockchain

A private blockchain, is nothing but managed blockchain, it is an authorized blockchain which is controlled by a single entity. In this blockchain a central authority decides who becomes a node.

336

S. A. Zambare and N. M. Sawant

Fig. 2 Types of blockchain

2.2.3

Consortium Blockchain

A permissioned blockchain is managed by a group of entities rather than a single entity like a private blockchain. Therefore, the consortium chain is more decentralized than the private chain, which provides a guarantee.

2.2.4

Hybrid Blockchain

It is a type blockchain and it is controlled by a single entity. It requires some degree of oversight over the public blockchain to perform validation of specific transactions (Fig. 2).

2.3 Consensus Mechanism 2.3.1

Proof of Work

In this example, miners mine blocks to solve cryptographic puzzles and connect them to the blockchain. This method requires a lot of energy and computing power. The puzzles are designed so that the system is difficult and demanding. When the miners solve the puzzle, the blocks are sent to the network for verification.

Performance Analysis of Patient Centric EHR Through Hyperledger Fabric

2.3.2

337

Proof of Stack

Proof of Stack is a blockchain method that works by selecting validators in proportion to their associated cryptocurrency holdings. This is done to avoid the computational cost of proof-of-work schemes.

2.3.3

RAFT

RAFT is a distributed, fault-based negotiation algorithm that allows the system to decide and serve the requestor in the event of a failure. RAFT has been approved by elected officials. Servers on Raft can be followers, leaders, or competitors. The leader is responsible for copying the diary to the followers [5].

3 Understanding Hyperledger Fabric in Detail Hyperledger Fabric is a blockchain model that provides the foundation for building blockchain products, solutions, and plug-and-play products designed for private businesses. The main difference between Hyperledger Fabric and other blockchain systems is that it is private and proprietary. The membership of the Hyperledger Fabric network is designed to represent an open, permission less system that allows unknown parties to join the network (processes such as Proof of Work must verify the work and security of the network and from the trusted service provider (MSP)) names. Below are the key architectures built into the Hyperledger Framework to build a complete and customizable blockchain business solution [8]. • Assets: The concept of an asset allows for the exchange of almost any value on the network, from natural foods to vintage cars and currencies of the future. • Chaincode: The implementation of Chaincode is separated from the exchange of orders, optimizing network scalability and performance by limiting the level of trust and proof required by different node types. • Ledger Features: Channel and personal data collection enable private and confidential multilateral transactions. This is typically required by competitors or regulated industries exchanging assets over a common network. • Confidentiality: The processing and collection of personal information ensures the privacy and confidentiality of transactions with multiple parties. This is often required by competitors or business management to share assets across the public network. • Enrollment Services and Security: Qualified members guarantee a secure blockchain network and members understand that all proceedings may be recognized and monitored by accredited regulators and auditors. • Consensus: a distinctive and consistent outlook gives your business flexibility and scalability.

338

S. A. Zambare and N. M. Sawant

3.1 Simple Hyperledger Fabric Network Three organizations Org1, Org2, and Org0, decide to create a network together. With CC1 configuration that matches all the peer organizations, CC1 is the Configuration Block for whole group of organization. Before starting communication all the organizations must create their identities of the admin from Certification Authority (CA), Every organization has their own Certification Authority (CA) here Org1, Org2 and Org0 having CA1, CA2 and CA0 are the Certification Authorities of these three Organizations [6] (Fig. 3). CA generate certificates required for respective node, administrators, organization definitions, and applications. It plays an important role in the network identifying organizational components. Peers are prime points in the network, so N/W Org1 and Org2, P1 and P2 are connected to the C1 environment and Org0 has O control of the environment. All nodes have a copy of the channel ledger (L1) where transactions are recorded. The judicial service keeps a copy of the registry, but not state records. Org1, Org2 also interacting with channels through their personal applications A1 and A2. Org1 and Org2 organizations join the channel. The application can access L1 record via S5 smart contract and create organization recognized transactions with the name specified in the authorization policy and listed on the form.

Fig. 3 Simple hyperledger fabric network

Performance Analysis of Patient Centric EHR Through Hyperledger Fabric

339

3.2 Type of Nodes Nodes are communication units in the blockchain. A “node” is just a single operating system and many different nodes can run on the same physical server. What matters is how nodes are grouped into “key domains” and how they join the organizations that manage them [7]. • Client or submitting-client: Customers send actual transaction calls to validators and trade proposals to the ordering service. • Peer: A client sends actual transaction calls to validators and trade proposals to an ordering service. • Ordering service: A service order that receives changes from different customers at the same time. The task of the order is to determine the batch of goods sent in an order and put them in blocks. These blocks are considered as blocks of the blockchain.

3.3 Smart Contract and Chaincode In Hyperledger Fabric, the blockchain code is a smart contract that runs on peers and generates transactions. Applications interact with blockchain ledgers through blockchain code. Therefore, the blockchain code must be installed on each peer-topeer application that supports the transaction and is initialized on the channel.

3.4 Benefits of Blockchain in EMR Keeping critical medical data safe is the most popular healthcare blockchain application today, and it’s no surprise. Safety is an important issue in medicine. Over 176 million medical records were compromised in data breaches between 2009 and 2017. Perpetrators stole credit card and bank details, health and genetic testing records. The main benefits of using blockchain on EMR are: • Decentralization: All stakeholders have access to the same copy of health records, and everyone has the same access and control. No single entity controls your data • Security and Privacy: Blockchain helps create a tamper-proof record due to its immutable properties. All records are encrypted, time-stamped and added chronologically to an existing distributed database. Patient privacy and identity are protected by cryptographic keys. • Data Ownership: Patients are assured that their records will not be misused or altered. He has full control over his data. This is achieved through strong cryptographic protocols and predefined smart contracts.

340

S. A. Zambare and N. M. Sawant

• Data Verifiability: Any interested party can verify the integrity and validity of the records stored on the blockchain. This feature is mainly applicable when a data validation process is required. • Trust: Blockchain records are accessible to anyone, so there is no need to question whether information has been tampered with for personal gain.

4 Implementation of Hyperledger Fabric Network • Admin dashboard page

• New Patient registration page

Performance Analysis of Patient Centric EHR Through Hyperledger Fabric

341

• EHR Docker

5 Conclusion Hyperledger Fabric is a promising blockchain framework with several concepts for providing secure identities that make policies, smart contracts, and records secure and controllable. This allows the EMR system to interoperate across multiple hospital organizations. Doctors can easily track your medical history. Patients will not need to carry a case file and digital records will be greatly improved. The EMR scenario falls into the category of private and closed blockchains, and the solution can successfully conclude that this is an encouraging structure for this type of blockchain. It provides a reliable and secure medical records management solution. Many improvements and modification can be made to improve the solution. Today, data is critical to the advancement of any technology. The model of machine learning contains lot of valid data and it needs to be re-authenticate. With this solution, you can safely examine a patient’s medical data without knowing the patient’s identity. This can improve the accuracy of machine learning models and will greatly improve the healthcare sector. I used text recipe. Physicians have to enter prescriptions into a patient’s EMR, which takes a lot of the doctor’s precious time. So it can be done by adding a voice recipe.

342

S. A. Zambare and N. M. Sawant

References 1. Alshalali, T., M’Bale, K., & Josyula, D. (2018). Security and privacy of electronic health records sharing using hyperledger fabric. In 2018 international conference on computational science and computational intelligence (CSCI), Las Vegas, NV, USA, 2018 (pp. 760–763). https://doi. org/10.1109/CSCI46756.2018.00152 2. Militello, C., La Iacona, S., Serbanati, L. D., Ricci, F. L., & Gilardi, M. C. (2015). A virtual health record-based EHR system. In 2015 E-health and bioengineering conference (EHB), Iasi, Romania, 2015 (pp. 1–6). https://doi.org/10.1109/EHB.2015.7391425 3. Pol, R. S., Deshmukh, A. B., Jadhav, M. M., Liyakat, K. K. S., & Mulani, A. O. (2022). iButton based physical access authorization and security system. Journal of Algebraic Statistics, 13(3), 3822–3829. 4. Kondekar, R., & Mulani, A. O. (2017). Raspberry pi based voice operated robot. International Journal of Recent Engineering Research and Development, 69–76 5. Kulkarni, P. R., Mulani, A. O., & Mane, P. B. (2017). Robust invisible watermarking for image authentication, Emerging trends in electrical, communications and information technologies. Springer. 6. Gajare, M., & Shedge, D. K. (2023). CMOS trans conductance based instrumentation amplifier for various biomedical signal analysis. International Journal of Intelligent Systems and Applications in Engineering, 11(3s), 63–71. 7. Mani, V., Manickam, P., Alotaibi, Y., Alghamdi, S., & Khalaf, O. I. (2021). Hyperledger healthchain: Patient-centric IPFS-based storage of health records. Electronics, 10, 3003. https:// doi.org/10.3390/electronics10233003 8. Dubovitskaya, A., Baig, F., Xu, Z., Shukla, R., Zambani, P. S., Swaminathan, A., Jahangir, M., Chowdhry, K., Lachhani, R., Idnani, N., Schumacher, M., Aberer, K., Stoller, S., Ryu, S., & Wang, F. (2020)., Wang F ACTION-EHR: Patient-centric blockchain-based electronic health record data management for cancer care. Journal of Medical Internet Research, 22(8), e13598. https://www.jmir.org/2020/8/e13598. https://doi.org/10.2196/13598