ADVANCES IN DATA SCIENCE AND

COMPUTING TECHNOLOGY

Methodology and Applications

AAP Research Notes on Optimization & Decision-Making Theories

ADVANCES IN DATA SCIENCE AND

COMPUTING TECHNOLOGY

Methodology and Applications

Edited by
Suman Ghosal, PhD

Amitava Choudhury, PhD

Vikram Kumar Saxena, MTech

Arindam Biswas, PhD

Prasenjit Chatterjee, PhD

First edition published 2023

Apple Academic Press Inc. CRC Press
1265 Goldenrod Circle, NE, 6000 Broken Sound Parkway NW,
Palm Bay, FL 32905 USA Suite 300, Boca Raton, FL 33487-2742 USA
760 Laurentian Drive, Unit 19, 4 Park Square, Milton Park,
Burlington, ON L7N 0A4, CANADA Abingdon, Oxon, OX14 4RN UK

© 2023 by Apple Academic Press, Inc.

Apple Academic Press exclusively co-publishes with CRC Press, an imprint of Taylor & Francis Group, LLC
Reasonable efforts have been made to publish reliable data and information, but the authors, editors, and publisher cannot assume
responsibility for the validity of all materials or the consequences of their use. The authors, editors, and publishers have attempted
to trace the copyright holders of all material reproduced in this publication and apologize to copyright holders if permission to
publish in this form has not been obtained. If any copyright material has not been acknowledged, please write and let us know so
we may rectify in any future reprint.
Except as permitted under U.S. Copyright Law, no part of this book may be reprinted, reproduced, transmitted, or utilized in any
form by any electronic, mechanical, or other means, now known or hereafter invented, including photocopying, microfilming, and
recording, or in any information storage or retrieval system, without written permission from the publishers.
For permission to photocopy or use material electronically from this work, access www.copyright.com or contact the Copyright
Clearance Center, Inc. (CCC), 222 Rosewood Drive, Danvers, MA 01923, 978-750-8400. For works that are not available on CCC
please contact [email protected]
Trademark notice: Product or corporate names may be trademarks or registered trademarks and are used only for identification
and explanation without intent to infringe.

Library and Archives Canada Cataloguing in Publication

Title: Advances in data science and computing technology : methodology and applications / edited by Suman Ghosal, PhD,
Amitava Choudhury, PhD, Vikram Kumar Saxena, MTech, Arindam Biswas, PhD, Prasenjit Chatterjee, PhD.
Names: Ghosal, Suman, editor. | Choudhury, Amitava, editor. | Saxena, Vikram Kumar, editor. | Biswas, Arindam, editor. |
Chatterjee, Prasenjit, 1982- editor.
Description: First edition. | Series statement: AAP research notes on optimization and decision-making theories | Includes
bibliographical references and index.
Identifiers: Canadiana (print) 20220220735 | Canadiana (ebook) 20220220808 | ISBN 9781774639979 (hardcover) |
ISBN 9781774639986 (softcover) | ISBN 9781003277071 (ebook)
Subjects: LCSH: Computer systems. | LCSH: Electronic data processing.
Classification: LCC QA76 .A38 2023 | DDC 004—dc23
Library of Congress Cataloging-in-Publication Data

CIP data on file with US Library of Congress

ISBN: 978-1-77463-997-9 (hbk)

ISBN: 978-1-77463-998-6 (pbk)
ISBN: 978-1-00327-707-1 (ebk)
AAP RESEARCH NOTES ON
OPTIMIZATION AND
DECISION-MAKING THEORIES

SERIES EDITORS:
Dr. Prasenjit Chatterjee
Department of Mechanical Engineering, MCKV Institute of Engineering,
Howrah, West Bengal, India
E-Mail: [email protected] / [email protected]
Dr. Dragan Pamucar
University of Defence, Military academy, Department of Logistics,
Belgrade, Serbia; E-Mail: [email protected]
Dr. Morteza Yazdani
Department of Business & Management, Universidad Loyola Andalucia,
Seville, Spain; E-Mail: [email protected]
Dr. Anjali Awasthi
Associate Professor and Graduate Program Director (M.Eng.), Concordia
Institute for Information Systems Engineering, Concordia, Canada
E-Mail: [email protected]

Most real-world search and optimization problems naturally involve multiple

criteria as objectives. Different solutions may produce trade-offs (conflicting
scenarios) among different objectives. A solution that is better with respect
to one objective may be a compromising one for other objectives. This
compels one to choose a solution that is optimal with respect to only one
objective. Due to such constraints, multi-objective optimization problems
(MOPs) are difficult to solve since the objectives usually conflict with each
other. It is usually hard to find an optimal solution that satisfies all objectives
from the mathematical point of view. In addition, it is quite common
that the criteria of real-world MOPs encompass uncertain information,
which becomes quite a challenging task for a decision maker to select the
criteria. Also, the complexities involved in designing mathematical models
increase. Considering, planning, and appropriate decision-making require
vi AAP Research Notes on Optimization and Decision-Making Theories

the use of analytical methods that examine trade-offs; consider multiple

scientific, political, economic, ecological, and social dimensions; and reduce
possible conflicts in an optimizing framework. Among all these, real-world
multi-criteria decision-making (MCDM) problems related to engineering
optimizations are categorically important and are quite often encountered
with a wide range of applicability.
MCDM problems are basically a fundamental issue in various fields,
including applied mathematics, computer science, engineering, management,
and operations research. MCDM models provide a useful way for modeling
various real-world problems and are extensively used in many different
types of systems, including, but not limited to, communications, mechanics,
electronics, manufacturing, business management, logistics, supply chain,
energy, urban development, waste management, and so forth.
In the aforementioned cases, modeling of multiple criteria problems
often becomes more complex if the associated parameters are uncertain and
imprecise in nature. Impreciseness or uncertainty exists within the param­
eters due to imperfect knowledge of information, measurement uncertainty,
sampling uncertainty, mathematical modeling uncertainty, etc. Theories like
probability theory, fuzzy set theory, type-2 fuzzy set theory, rough set, grey
theory, neutrosophic uncertainty theory available in the existing literature
deal with such uncertainties. Nevertheless, the uncertain multi-criteria
characteristics in such problems are not explored in depth, and a lot can be
achieved in this direction. Hence, different mathematical models of real-life
multi-criteria optimization problems can be developed on various uncer­
tain frameworks with special emphasis on sustainability, manufacturing,
communications, biomedical, electronics, materials, energy, agriculture,
environmental engineering, strategic management, flood risk management,
supply chain, waste management, transportations, economics, and industrial
engineering problems, to name a few.
Coverage & Approach:
The primary endeavor of this series is to introduce and explore contemporary
research developments in a variety of rapidly growing decision-making
areas. The volumes will deal with the following topics:
• Crisp MCDM models
• Rough set theory in MCDM
• Fuzzy MCDM
• Neutrosophic MCDM models
• Grey set theory
AAP Research Notes on Optimization and Decision-Making Theories vii

• Mathematical programming in MCDM

• Big data in MCDM
• Soft computing techniques
• Modelling in engineering applications
• Modeling in economic issues
• Waste management
• Agricultural practice
• Material selection
• Renewable energy planning
• Industry 4.0
• Sustainability
• Supply chain management
• Environmental policies
• Manufacturing processes planning
• Transportation and logistics
• Strategic management
• Natural resource management
• Biomedical applications
• Future studies and technology foresight
• MCDM in governance and planning
• MCDM and social issues
• MCDM in flood risk management
• New trends in multi-criteria evaluation
• Multi-criteria analysis in circular economy
• Multi-criteria evaluation for urban and regional planning
• Integrated MCDM approaches for modeling relevant applications
and real-life problems

Types of volumes:
This series reports on current trends and advances in optimization and
decision-making theories in a wider range of domains for academic and
research institutes along with industrial organizations. The series will cover
the following types of volumes:
• Authored volumes
• Edited volumes
• Conference proceedings
• Short research (thesis-based) books
• Monographs
viii AAP Research Notes on Optimization and Decision-Making Theories

Features of the volumes will include recent trends, model extensions,

developments, real-time examples, case studies, and applications. The
volumes aim to serve as valuable resources for undergraduate, postgraduate
and doctoral students, as well as for researchers and professionals working
in a wider range of areas.

CURRENT & FORTHCOMING BOOKS IN THE SERIES

Multi-Criteria Decision-Making Techniques in Waste Management:
A Case Study of India
Suchismita Satapathy, Debesh Mishra, and Prasanjit Chatterje
Applications of Artificial Intelligence in Business and Finance:
Modern Trends
Editors: Vikas Garg, Shalini Aggarwal, Pooja Tiwari, and Prasenjit Chatterjee
Advances in Data Science and Computing Technology:
Methodology and Applications
Editors: Suman Ghosal, Amitava Choudhury, Vikram Kr. Saxena, Arindam Biswas,
and Prasenjit Chatterjee
Machine Learning Algorithms in Security Analytics: Applications,
Principles, and Practices
Editors: Karan Singh, Latha Banda, and Manisha Manjul
Sustainable Development: A Geospatial and Statistical Perspective
Editors: Qazi Mazhar Ali, Rizwan Ahmad, Irfan Ali, Prasenjit Chatterjee, and
Harjit Pal Singh
Decision Analysis and Optimization Models for Sustainable
Development Goals 2030
Irfan Ali, Umar Muhammad Modibbo, and Bui Thanh Hung
Smart Sensors, Actuators, and Decision Support Systems for Precision
Agriculture
Editors: Narendra Khatri, Ajay Kumar Vyas, Celestine Iwendi, and Prasenjit
Chatterjee
Attacks on Artificial Intelligence: The New Facets of Cyber Ecospace
Editors: Kukatlapalli Pradeep Kumar, Vinay Jha Pillai, and Boppuru Rudra Prathap
ABOUT THE EDITORS

Suman Ghosal, PhD

Assistant Professor and Head of the Department of Electronics,
Pailan College of Management and Technology, West Bengal, India,
E-mail: [email protected]
Suman Ghosal, PhD, is Assistant Professor and Head of the Department of
Electrical and Electronics Engineering at Pailan College of Management
and Technology, Kolkata, India. He is also Administrative In-Charge of
Pailan Technical Campus, Kolkata. He has more than 12 years of academic
experience in addition to industrial experience. His broad areas of interest
include renewable energy, electronics devices, and soft computing. He has
published many research papers in various national and international reputed
journals and conferences. He obtained his BTech degree from Maulana Abul
Kalam Azad University of Technology (formerly WBUT), Kolkata, and his
MTech from Bengal Engineering and Science University, Shibpur. He is
currently pursuing a PhD at the Indian Institute of Engineering Science and
Technology, Shibpu, India.

Amitava Choudhury, PhD

Assistant Professor, School of Computer Science, University of Petroleum
and Energy Studies, Dehradun, Uttarakhand, India
Amitava Choudhury, PhD, is Assistant Professor in the School of Computer
Science, University of Petroleum and Energy Studies, Dehradun, India. He
has more than ten years of experience in teaching and many years in research
work. He serves as a reviewer for IEEE Biomedical Transaction and the
journal Medical & Biological Engineering & Computing. He is member of
IEEE, UP section, and the Institute of Engineers, India. His area of research
interest is computational geometry in the field of micromechanical modelling,
pattern recognition, character recognition, and machine learning. Currently,
he is associated with the journal Cluster Computing as a guest editor. He
earned his MTech from Jadavpur University and submitted his PhD thesis at
the Indian Institute of Engineering Science and Technology, Shibpur, India.
x About the Editors

Vikram Kumar Saxena, MTech

Assistant Professor, Pailan College of Management and Technology,
West Bengal, India
Vikram Kumar Saxena is Assistant Professor at the Pailan College of Manage­
ment and Technology Kolkata, India. He has been a teaching professional
for more than eight years. His research area of interest includes electrical
machines, power systems, and power electronics. He has published many
research papers in national and international journals and conferences. He
received his BTech from West Bengal University of Technology, Kolkata,
and his MTech from the National Institute of Technology Patna, India.

Arindam Biswas, PhD

Associate Professor, School of Mines and Metallurgy,
Kazi Nazrul University, West Bengal–713305, India,
E-mail: [email protected]
Arindam Biswas, PhD, is Associate Professor in the School of Mines and
Metallurgy at Kazi Nazrul University, Asansol, West Bengal, India. Dr. Biswas
has 10 years of experience in teaching research and administration. He has
published many technical papers in journals and conference proceedings as
well as six books, an edited volume, and a book chapter. Dr. Biswas was
awarded several research grants, such as from the Science and Engineering
Research Board, Govt of India, and the Centre of Biomedical Engineering,
Tokyo Medical and Dental University, in association with RIE, Shizouka
University, Japan. Presently Dr. Biswas is serving as an Associate Editor
of the journal Cluster Computing and as a guest editor of Nanoscience and
Nanotechnology and Recent Patents on Materials Science. He was formerly a
Visiting Professor at the Research Institute of Electronics, Shizouka University,
Japan. He has organized and chaired several international conferences in India
and abroad. Dr. Biswas acted as reviewer for journals and is a member of
the Institute of Engineers (India) and regular fellow of Optical Society of
India (India). His research interest is in carrier transport in low-dimensional
systems and electronic devices, non-linear optical communications, and THz
Semi-conductor sources. Dr. Biswas received an MTech degree in Radio
Physics and Electronics from the University of Calcutta, India, and a PhD
from NIT Durgapur, India. He was a postdoctoral tesearcher at Pusan National
University, South Korea, with prestigious BK21PLUS Fellowship, Republic
of Korea.
About the Editors xi

Prasenjit Chatterjee, PhD

Dean (Research and Consultancy) MCKV Institute of Engineering,
West Bengal, India
Prasenjit Chatterjee, PhD, is the Dean (Research and Consultancy) at MCKV
Institute of Engineering, India. He has published over 90 research papers
in international journals and peer-reviewed conferences. He has received
numerous awards, including Best Track Paper Award, Outstanding Reviewer
Award, Best Paper Award, Outstanding Researcher Award, and University
Gold Medal. He has been the guest editor of several special issues and has
authored and edited several books on decision-making approaches, supply
chains, and sustainability modeling. He is the lead editor of the book series
AAP Research Notes on Optimization and Decision-Making Theories and
Frontiers of Mechanical and Industrial Engineering with Apple Academic
Press. Dr. Chatterjee is one of the developers of two “multiple-criteria
decision-making methods called measurement of alternatives and ranking
according to compromise solution” (MARCOS) and “ranking of alternatives
through functional mapping of criterion sub-intervals into a single interval”
(RAFSI).
CONTENTS

Contributors......................................................................................................... xvii

Abbreviations ...................................................................................................... xxiii

Preface ............................................................................................................... xxvii

Part I: Recent Application of Deep Learning and ANN Paradigm ................... 1

1. Deep Learning Supported Evaluation of Retinal Vessel

Enhancement Techniques.............................................................................. 3

Mahua Nandy Pal, Minakshi Banerjee, and Shuvankar Roy

2. An Efficient and Optimized Deep Neural Network Architecture

for Handwritten Digit Recognition............................................................. 19

Shuvankar Roy and Mahua Nandy Pal

3. Design and Development of a Driving Assistance and

Safety System Using Deep Learning .......................................................... 35

Shekhar Verma, Tushar Bajaj, Nidhi Sindhwani, and Abhijit Kumar

4. Prediction of the Positive Cases of COVID-19 for Individual States

in India and Its Measure: A State-of-the-Art Analysis ............................. 47

Amitava Choudhury, Kaushik Ghosh, and Suman Ghosal

5. Deep Residual Neural Network Based Multi-Class Image

Classifier for Tiny Images ........................................................................... 59

Alok Negi, Prachi Chauhan, and Amitava Choudhury

6. Improved TDOA-Based Node Localization in WSN:

An ANN Approach ....................................................................................... 69

Arpan Chatterjee, Ashish Agarowala, and Parag Kumar Guha Thakurta

Part II: Role and Impact of Big Data in E-Commerce and the

Retail Sector.................................................................................................. 81

7. Impact of Big Data on E-Commerce Websites .......................................... 83

Divyang Bhartia and Mausumi Das Nath

8. A Review of the Revolutionizing Role of Big Data in Retail Industry .... 95

Mausumi Das Nath and Madhu Agarwal Agnihotri

xiv Contents

Part III: Algorithm for Load Balancing in Cloud Computing ...................... 107

9. Load-Balancing Implementation with an Algorithm of

Matrix Based on the Entire Jobs .............................................................. 109

Payel Ray, Enakshmi Nandi, Ranjan Kumar Mondal, and Debabrata Sarddar

10. Load-Balancing Algorithm with Overall Tasks....................................... 119

Enakshmi Nandi, Payel Ray, Ranjan Kumar Mondal, and Debabrata Sarddar

11. Implementation of Load Balancing of Matrix Problem with the

Row Sum Method....................................................................................... 131

Payel Ray, Enakshmi Nandi, Ranjan Kumar Mondal, and Debabrata Sarddar

Part IV: Advances in Embedded System-Based Applications ....................... 143

12. Microcontroller-Based Wearable Location Tracker System

with Immediate Support Facility.............................................................. 145

Sumanta Chatterjee, Dwaipayan Saha, Indrani Mukherjee, and Jesmin Roy

13. Design and Modifications of Eyedrop Bottle Holder Device:

A Simplest Way to Install Ocular Drug Independently.......................... 169

Sirajum Monira, Vishal Biswas, and Satyam Jha

14. Agro-Nutrition Alert (ANA)...................................................................... 187

Rajneesh Kumar, Priya Anand, Debaparna Sengupta, and Monalisa Datta

Part V: Optimization Technique Using MATLAB Platform.......................... 195

15. Selection of Optimal Mother Wavelet for Fault Analysis in

Induction Motor Using Stator Current Waveform................................. 197

Arunava Kabiraj Thakur, Palash Kumar Kundu, and Arabinda Das

16. High Efficiency High Gain SEPIC-Buck Boost Converter Based

BLDC Motor Drive for Solar PV Array Fed Water Pumping ............... 219

K. M. Ashitha and D. Thomas

17. A Classy Fuzzy MPPT Controller for Standalone PV System............... 237

Rahul Kumar Singh, Rahul Kumar Prasad, Rahul Singh, Aayush Ashish,

Debaparna Sengupta, Rishiraj Sarker, and Monalisa Datta

18. Active and Reactive Power Control in Variable Wind Turbine

Using Simultaneous Working of STATCOM and Pitch Control ........... 253

Arrik Khanna, Pushpanjali Singh Bisht, and Vikram Kumar Saxena

Part VI: Technique for Improving Information and Network Security ....... 273

19. Information Security Using Key Management ....................................... 275

Sanjib Halder, Bijoy Kumar Mandal, and Arindam Biswas

Contents xv

20. Strength-Based Novel Technique for Malicious Nodes Isolation........... 289

Prachi Chauhan and Alok Negi

21. Securing the Information Using Combined Method .............................. 305

Bijoy Kumar Mandal, Saptarshi Roychowdhury, Payel Majumder, and Anwesa Das

22. Implementation of DNA Cryptography in IoT Using Chinese

Remainder Theorem, Arithmetic Encoding, and Asymmetric

Key Cryptography ..................................................................................... 319

Ananya Satpati (Das), Soumya Paul, and Payel Majumder

23. Robust Technique to Overcome Thwarting Communication

Under Narrowband Jamming in Extremely Low Frequency of

CDMA-DSSS System on Coding- Based MATLAB Platform ............... 333

Tanajit Manna, Rahit Basak, Mainak Das, and Alok Kole

Part VII: Miscellaneous Topics in Computing ................................................ 347

24. The Utility of Regional Language Facility in Accounting

Software for Retail Market of West Bengal............................................. 349

Madhu Agarwal Agnihotri and Arkajyoti Pandit

25. Comparative Analysis of Storey Drift of a G+9 RC-Structured

Building Due to Seismic Loads in Various Seismic Regions .................. 363

Budhaditya Dutta, Aditya Narayan Chakraborty, Ayushi Shah, Debarka Brahma,

Amit Deb, and Tushar Kanti De

26. A Proposal for Disinfection of Contaminated Subject Area

Through Light ............................................................................................ 383

Anubrata Mondal and Kamalika Ghosh

Index .................................................................................................................... 397

CONTRIBUTORS

Ashish Agarowala
CSE Department, NIT Durgapur, West Bengal, India

Madhu Agarwal Agnihotri

Assistant Professor, Department of Commerce, St. Xavier’s College (Autonomous), Kolkata, West Bengal,
India, E-mails: [email protected]; [email protected]

Priya Anand
Department of Information Science and Engineering, AMC Engineering College, Bangalore–560083,
Karnataka, India

Aayush Ashish
Techno International New Town, Kolkata–700156, West Bengal, India

K. M. Ashitha
Department of Electrical and Electronics Engineering, Amal Jyothi College of Engineering, Kanjirappally,
Kerala–686518, India, E-mail: [email protected]

Tushar Bajaj
Department of ECE, Amity School of Engineering and Technology, New Delhi GGSIPU, Delhi, India

Minakshi Banerjee
Computer Science and Engineering Department, RCC Institute of Information Technology,
Kolkata–700015, West Bengal, India
Rahit Basak
Computer Science and Engineering Department, Pailan College of Management and Technology,
Kolkata–700104, West Bengal, India

Divyang Bhartia
Student, St. Xavier’s College (Autonomous), Kolkata, 30 Mother Teresa Sarani, Kolkata–700016,
West Bengal, India

Pushpanjali Singh Bisht

Department of Electrical Engineering, DPGITM Gurugram, Haryana–122001, India

Arindam Biswas
Kazi Nazrul University, Asansol, West Bengal, India

Vishal Biswas
Department of Bachelor in Optometry, Pailan College of Management and Technology, Joka,
Kolkata–700104, West Bengal, India

Debarka Brahma
Undergraduate Student, Department of Civil Engineering, Amity University Kolkata, Kolkata,
West Bengal–700135, India

Aditya Narayan Chakraborty

Undergraduate Student, Department of Civil Engineering, Amity University Kolkata, Kolkata,
West Bengal–700135, India
xviii Contributors

Arpan Chatterjee
CSE Department, NIT Durgapur, West Bengal, India

Sumanta Chatterjee
Department of Computer Science and Engineering, JIS College of Engineering, Kalyani, West Bengal,
India, E-mail: [email protected]
Prachi Chauhan
G.B. Pant University of Agriculture and Technology, Pantnagar, Uttarakhand, India,
E-mail:[email protected]
Amitava Choudhury
School of Computer Science, University of Petroleum and Energy Studies, Dehradun, Uttarakhand,
India, E-mail: [email protected]
Anwesa Das
NSHM Knowledge Campus, Durgapur, West Bengal, India

Arabinda Das
Department of Electrical Engineering, Jadavpur University, Kolkata–700032, West Bengal, India

Mainak Das
Computer Science and Engineering Department, Pailan College of Management and Technology,
Kolkata–700104, West Bengal, India

Monalisa Datta
Department of Electrical Engineering, Techno International New Town, Kolkata–700156, West Bengal,
India

Tushar Kanti De
Associate Professor and Head of Department, Department of Civil Engineering,
Budge Budge Institute of Technology, Budge Budge, West Bengal–700137, India
Amit Deb
Assistant Professor, Department of Civil Engineering, Amity University Kolkata, Kolkata,
West Bengal–700135, India

Budhaditya Dutta
Undergraduate Student, Department of Civil Engineering, Amity University Kolkata, Kolkata,
West Bengal–700135, India, E-mail: bdneil3103@gmail

Suman Ghosal
Department of Electrical and Electronics Engineering, Pailan College of Management and Technology,
Kolkata, West Bengal, India

Kamalika Ghosh
Director, School of Illumination Science, Engineering and Design, Jadavpur University, Kolkata,
West Bengal, India, E-mail: [email protected]

Kaushik Ghosh
School of Computer Science, University of Petroleum and Energy Studies, Dehradun, Uttarakhand, India

Sanjib Halder
The Bhawanipur Education Society College, Kolkata, West Bengal, India

Satyam Jha
Department of Bachelor in Optometry, Pailan College of Management and Technology, Joka,
Kolkata–700104, West Bengal, India
Contributors xix

Arrik Khanna
Department of Electrical Engineering, Chitkara University, Punjab–140401, India,
E-mail: [email protected]

Alok Kole
Electrical Engineering Department, RCC Institute of Information Technology, Kolkata–700015,
West Bengal, India
Abhijit Kumar
School of Computer Science, University of Petroleum and Energy Studies, Dehradun, Uttarakhand,
India, E-mail: [email protected]

Rajneesh Kumar
Department of Electrical Engineering, Techno International New Town, Kolkata–700156, West Bengal,
India, E-mail: [email protected]

Palash Kumar Kundu

Department of Electrical Engineering, Jadavpur University, Kolkata–700032, West Bengal, India

Payel Majumder
Department of Computer Science and Engineering, NSHM Knowledge Campus, Durgapur,
Arrah Shibtala Via Muchipara, Durgapur–713212, West Bengal, India

Bijoy Kumar Mandal

NSHM Knowledge Campus, Durgapur, West Bengal, India, E-mail: [email protected]

Tanajit Manna
Electronics and Communication Engineering Department, Pailan College of Management and
Technology, Kolkata–700104, West Bengal, India, E-mail: [email protected]

Anubrata Mondal
Research Scholar, School of Illumination Science Engineering and Design,

(Jadavpur University, Kolkata) and Assistant Professor, Department of Electrical Engineering,

Global Institute of Management and Technology, West Bengal, India

Ranjan Kumar Mondal

Department of Computer Science and Engineering, University of Kalyani, West Bengal, India,
E-mail: [email protected]

Sirajum Monira
Department of Bachelor in Optometry, Pailan College of Management and Technology, Joka,
Kolkata–700104, West Bengal, India, E-mail: [email protected]

Indrani Mukherjee
Department of Computer Science and Engineering, JIS College of Engineering, Kalyani, West Bengal,
India

Enakshmi Nandi
Department of Computer Science and Engineering, University of Kalyani, West Bengal, India,
E-mail: [email protected]

Mausumi Das Nath

Assistant Professor, Department of Commerce, St. Xavier’s College (Autonomous), Kolkata,
30 Mother Teresa Sarani, Kolkata – 700016, West Bengal, India, E-mail: [email protected]

Alok Negi
National Institute of Technology, Uttarakhand, India
xx Contributors

Mahua Nandy Pal

Computer Science and Engineering Department, MCKV Institute of Engineering, Howrah–711204,
West Bengal, India, E-mail: [email protected]

Arkajyoti Pandit
Assistant Professor, Department of Commerce, St. Xavier’s College (Autonomous), Kolkata,
West Bengal, India
Soumya Paul
Department of Computer Science and Engineering, NSHM Knowledge Campus, Durgapur,
Arrah Shibtala Via Muchipara, Durgapur–713212, West Bengal, India

Rahul Kumar Prasad

Techno International New Town, Kolkata–700156, West Bengal, India

Payel Ray
Department of Computer Science and Engineering, University of Kalyani, West Bengal, India,
E-mail: [email protected]

Jesmin Roy
Department of Computer Science and Engineering, JIS College of Engineering, Kalyani, West Bengal,
India

Shuvankar Roy
Computer Science and Engineering Department, MCKV Institute of Engineering, Howrah–711204,
West Bengal, India, E-mail: [email protected]

Saptarshi Roychowdhury
NSHM Knowledge Campus, Durgapur, West Bengal, India

Dwaipayan Saha
Department of Computer Science and Engineering, JIS College of Engineering, Kalyani, West Bengal,
India

Debabrata Sarddar
Department of Computer Science and Engineering, University of Kalyani, West Bengal, India,
E-mail: [email protected]

Rishiraj Sarker
Jadavpur University, Kolkata–700032, West Bengal, India, E-mail: [email protected]

Ananya Satpati (Das)

Department of Computer Science and Engineering, NSHM Knowledge Campus, Durgapur, Arrah
Shibtala Via Muchipara, Durgapur–713212, West Bengal, India, E-mail: [email protected]

Vikram Kumar Saxena

Department of Electrical Electronics and Engineering, Pailan College of Management and Technology,
West Bengal–700102, India

Debaparna Sengupta
Department of Electrical Engineering, Techno International New Town, Kolkata–700156, West Bengal,
India

Ayushi Shah
Undergraduate Student, Department of Civil Engineering, Amity University Kolkata, Kolkata,
West Bengal–700135, India

Nidhi Sindhwani
Department of ECE, Amity School of Engineering and Technology, New Delhi GGSIPU, Delhi, India
Contributors xxi

Rahul Kumar Singh

Techno International New Town, Kolkata–700156, West Bengal, India

Rahul Singh
Techno International New Town, Kolkata–700156, West Bengal, India

Arunava Kabiraj Thakur

Department of Electrical Engineering, Techno Main Salt Lake, Sector V, Bidhannagar, Kolkata–700091,
West Bengal, India, E-mail: [email protected]

Parag Kumar Guha Thakurta

CSE Department, NIT Durgapur, West Bengal, India

D. Thomas
Amal Jyothi College of Engineering, Kanjirappally, Kottayam, Kerala–686518, India

Shekhar Verma
Department of ECE, Amity School of Engineering and Technology, New Delhi GGSIPU, Delhi, India
ABBREVIATIONS

2D two dimensional
ACC adaptive cruise control
AGC adaptive gamma correction
AHE adaptive histogram equalization
ANA agro nutrition alert
AWGN additive white gaussian noise
BER bit error rate
BS base station
BT body temperature
CDF cumulative distribution function
CDMA code division multiple access
CLAHE contrast limited adaptive histogram equalization
CNN/ConvNet convolutional neural network
COG center of gravity
Coif coiflets
CQA classification quality assessment
CSA control based pitch actuator system
CTC connectionist temporal classification
db Daubechies
DES data encryption standard
DGR directional greedy routing
DL dead load
DNA deoxyribonucleic acid
DNN deep neural network
DSSS direct sequence spread spectrum
DWT discrete wavelet transform
EBGR edge node based greedy routing
ESU energy storage unit
EVCSs electric vehicle charging stations
EVs electric vehicles
FFT fast Fourier transform
FLCs fuzzy logic controllers
FN false negative
FP false positive
xxiv Abbreviations

GPS global positioning system

GSM global system for mobile communications
GWEC Global Wind Energy Council
GyTAR greedy traffic-aware routing
HOG histograms of oriented gradients
HR heart rate
HRV heart rate variability
HUD heads up display
I/O input/output
ICP ideal customer profile
IDWT inverse discrete wavelet transform
IoT internet of things
IQA image quality assessment
IR infrared
IRB institutional review board
KNN k-nearest neighbor
LBMM load balance Min-Min
LED light emitting diode
LiDAR light detection and ranging
LL live load
LSTM long short term memory
LVRT low voltage ride through
MCT minimum completion time
MM min-min
MNIST modified national institute of standards and technology
MOM mean of maximu m
MPPT maximum power point tracking
MSE mean squared error
NB negative big
NS negative small
OLB opportunistic load balancing
P&O perturb and observe
PB positive big
PS positive small
PSD power spectral density
PSNR peak signal to noise ratio
ReLU rectified linear unit
RMSE root of mean square error
RNA ribonucleic acid
Abbreviations xxv

RNN recurrent neural network

ROI region of interest
RSS received signal strength
RSU road side unit
SD standard deviation
SGD stochastic gradient descent
SGDM stochastic momentum gradient descent
SISED School of Illumination Science and Engineering
Sn sensitivity
STATCOM static compensator
STC standard test conditions
STFT short term Fourier transform
SVM support vector machine
Sym Symlets
TDOA time difference of arrival
TN true negative
TP true positive
TSO transmission system handling operators
UV ultraviolet
V voltage
V2V vehicle to vehicle
VSI voltage switching inverters
VUV vacuum UV
WD weighting distribution
WHO World Health Organization
WSN wireless sensor network
ZE zero
PREFACE

This book aims to provide a sharing platform for researchers, academicians,

and industry to throw in their expertise, acquaintance, and understanding in
a wider domain of progressive and emerging technologies, including arti­
ficial intelligence, communication, cyber security, data analytics, Internet
of Things (IoT), machine learning, power system, VLSI, embedded system,
and much more. It aims to be a transdisciplinary and cross-cutting resource
that present recent trends and advances and challenges.
We are delighted to deliver this book with the support and contributions
from global academicians and researchers. This book is a manifestation of
various interesting and important aspects of data science and computing
technologies and their applications and methodologies in a wider range
of applications including deep learning, DNA cryptography, classy fuzzy
MPPT controller, driving assistance, and safety systems.
This book consists of 26 chapters, covering discussion on innovative
algorithms and their applications for solving cutting-edge computational and
data science problems with an interdisciplinary research perspective. All 26
chapters are divided in following seven parts containing interdisciplinary
applications and methodologies of data science and computing technologies:
 Part I: Recent Application of Deep Learning and ANN Paradigm
 Part II: Role and Impact of Big Data in E-Commerce and Retail Sector
 Part III: Algorithm for Load Balancing in Cloud Computing
 Part IV: Advances in Embedded System Based Application
 Part V: Optimization Technique Using MATLAB Platform
 Part VI: Technique for Improving Information and Network Security
 Part VII: Miscellaneous Topics in Computing
In summary, in this 21st century, there is a genuine need for technology
upgradation, and this book provides interdisciplinary research and a techno­
logical overview that the editors hope will lead to open a new dimension of
understanding, which will be useful for uncovering new in this growing area.
The editors of this book are thankful to all authors, reviewers, and Apple
Academic Press for their valuable input and contributions to make this book
possible.
—Editors
 PART I

Recent Application of Deep Learning

and ANN Paradigm

CHAPTER 1

DEEP LEARNING SUPPORTED

EVALUATION OF RETINAL VESSEL
ENHANCEMENT TECHNIQUES
MAHUA NANDY PAL,1 MINAKSHI BANERJEE,2 and
SHUVANKAR ROY1
1
Computer Science and Engineering Department,

MCKV Institute of Engineering, Howrah–711204, West Bengal, India,

E-mail: [email protected] (M. N. Pal)

2
Computer Science and Engineering Department, RCC Institute of

Information Technology, Kolkata–700015, West Bengal, India

ABSTRACT

Hypertension, diabetes, glaucoma, etc., are different diseases, the symptoms

of which prevails in retinal fundus images at preliminary stage. Retinal
vessel segmentation is a significant prior step for different automated disease
analysis system of retinal fundus images. Further, image enhancement has a
fundamental role in retinal vessel segmentation. The better is the enhance­
ment the better segmentation result is expected. In this work, three enhance­
ment techniques are considered which are very frequently used before retinal
vessel segmentation in contemporary literatures. The competency of each of
the enhancement techniques is evaluated using two different categories of
image quality assessment (IQA) metrics, peak signal to noise ratio (PSNR)
and absolute mean brightness error (AMBE). In this way, morphological
operation performs the best when compared with other two enhancement
techniques. This observation has been further evaluated from a reverse view

Advances in Data Science and Computing Technology: Methodology and Applications. Suman Ghosal,
Amitava Choudhury, Vikram Kumar Saxena, Arindam Biswas, & Prasenjit Chatterjee (Eds.)
© 2023 Apple Academic Press, Inc. Co-published with CRC Press (Taylor & Francis)
4 Advances in Data Science and Computing Technology

point of effective deep convolutional neural network (CNN) vessel classifi­

cation result on three differently enhanced image dataset which supports the
previous observation in terms of classification efficiency metrics as well.
Better vessel classification metrics are obtained from the same classification
model when input datasets are better enhanced to represent vessel.

1.1 INTRODUCTION

Retinal fundus image is characterized both by anatomical and pathological

artifacts present in its surface. The key anatomical artifacts are blood vessels,
optic disc, and fovea. Pathological artifacts are usually symptoms of different
optical and cardio-vascular diseases. These are manifested at an early stage
in retinal image surface. Automated retinal image analysis system helps
medical professionals in fast screening of huge real life dataset. Challenges
present in automated retinal diagnostic system are presence of noise, low
contrast, uneven illumination, minute sizes of medically significant local
artifacts such as micro aneurysms, hemorrhages, etc. In process of automated
retinal analysis system development, retinal vessel segmentation plays a
significant role, Moreover, retinal image enhancement is the pre requisite of
retinal vessel segmentation.

1.2 LITERATURE SURVEY

Adaptive histogram equalization (AHE) is an image enhancement

technique. But AHE reports two major problems namely slow speed
and over enhancement of noise. These problems are addressed by Pizer
et al. (1987). Zuiderveld in 1994 presented contrast limited adaptive
histogram equalization (CLAHE), to acquire required results overcoming
the disadvantages Nandy and Banerjee (2017) made use of CLAHE for
enhancement of images before vessel segmentation. Huang, Cheng, and
Chiu (2013) dealt with AGC based image enhancement. They presented
image transformation through gamma correction and probability distribution
of luminance pixels. Rahman et al. (2016) discussed application of AGC
for different types of image enhancement. An adaptive gamma correction
(AGC) based on image contrast is adopted in Nandy and Banerjee (2019).
Application of morphological operation has been utilized in medical image
enhancement in 2015 (Hassanpour et al., 2015). Hassan et al. (2015) used
different morphological operators and clustering for vessel segmentation. An
Deep Learning Supported Evaluation of Retinal Vessel 5

entropy based image quality metric has been worked out in 2012 (Nandy
and Banerjee, 2012). Sheba and Gladston (2016) evaluated different image
enhancement techniques in the domain of mammogram images.
Almotiri et al. (2018) presented a survey on different retinal vessel
segmentation algorithms. Roychowdhury et al. (2015) iteratively thresholded
the residual retinal image to obtain higher vessel segmentation accuracy. Fu
et al. (2016); Liskowski et al. (2016); and Mo and Zhang (2017) exploited
deep learning architecture to identify vessels from retinal fundus images. Fu
et al. (2016) used convolutional neural network (CNN) which follows vessel
probability map. In this chapter, fully connected conditional random fields
are combined with probability map to segment vessels. Liskowski et al.
(2016) applied several preprocessing steps and training was conducted on a
large number of samples. Generally large volume of training data is required
for effective learning of deep architecture. These processes are extremely
computationally exhaustive and the training is also very time consuming
with requirements of expensive resources.
Ronneberger et al. (2015) proposed U-net CNN model which is efficient
in segmenting bio medical images.
The contribution is to find the effective enhancement technique quantita­
tively following two diverse viewpoints. We have experimented with some
of the enhancement techniques, frequently used for retinal vessel segmenta­
tion such as, CLAHE, AGC and morphological enhancement techniques. On
the basis of IQA metric values, we have established with the notion that
morphological enhancements are the best among them in an application
oriented way. To arrive at a conclusive result from a different dimension, we
experimented with the efficiency of deep vessel classification on three differ­
ently enhanced image datasets also. In this phase, the better the classification
evaluation metric values the more informative the input images are. The
results obtained from deep classification experiments are decently aligned
with the observations obtained from image quality assessment (IQA) too.
The work representation is structured as follows. Section 1.3 presents
enhancement methods normally used before vessel segmentation of retinal
fundus. IQA evaluation metrics are there in Section 1.4 and deep classifica­
tion based evaluation metrics. Section 1.5 presents database description.
Section 1.6 explores deep neural network (DNN) architecture, training
details and implementation requirements used to evaluate our observation.
Section 1.7 describes experimental results of IQA based evaluation and deep
CNN classification based evaluation respectively. Section 1.8 demonstrates
the experimental results found in Section 1.7. Section 1.9 is the conclusion
of the work.
6 Advances in Data Science and Computing Technology

1.3 IMAGE ENHANCEMENT TECHNIQUES

1.3.1 CONTRAST LIMITED ADAPTIVE HISTOGRAM-BASED

ENHANCEMENT

General image transformation function is:

=s T ( r ) ; 0 ≤ r ≤ lmax (1)
The lmax is maximum intensity value.
Pixel intensity levels of an image may be thought of as random variables.
Random variables are fundamentally described by probability density func­
tion. Thus, the transformation function is:
r
= (r )
s T= ∫0
pr ( w ) dw (2)
where; w is a dummy integration variable. Eqn. (2) represents the cumulative
distribution function (CDF) (digital image processing, Gonzalez). Following
discrete representation, the probabilities of occurrence of intensity level rk is:
nk
Pr ( rk ) = (3)
MN
where; nk is the number of pixels having intensity level rk. So, the discrete
variation of the function is:
( lmax )
= ( rk )
sk T= (4)
MN ∑ j =0n j
k

where; k = 0, 1, …, lmax
Eqn. (4) represents histogram equalized image. Pizer et al. (1987) elabo­
rated AHE application. Zuiderveld (1994) modified it to CLAHE to obtain
better enhancement result. CLAHE follows local region based approach. All
the local regions contrast enhancement takes place independently and are
combined with bilinear interpolation. Thus, boundary effects between the
region can be removed.

1.3.2 ADAPTIVE GAMMA CORRECTION (AGC)-BASED

ENHANCEMENT

AGC with weighting distribution (WD) (Huang, Cheng, and Chiu, 2013) is
depicted as follows. Gamma correction formula is:
Deep Learning Supported Evaluation of Retinal Vessel 7

γ
 l 
T ( lmax ) = ( lmax )   (5)
 lmax 
where; lmax is the maximum intensity. The gamma value may be greater than
1 or may be less than 1 (Huang et al., 2013). AGC with weight distribution
is as follows (Huang et al., 2013):
1−cdf ( l )
T ( l ) =lmax ( l / lmax ) (6)
The underlined concept is that the process tries to increase lower intensity
level and at the same time tries to resist decrease of higher intensity level.
The WD function has the following formula:
pdf ( l ) − pdf min
pdf w ( l ) = pdf max ( )a (7)
pdf max − pdf min
where; a is the adjusted parameter. Thus, cdf becomes:
pdf w ( l )
cdf w ( l ) = ∑ lmax
l
(8)
=0
∑ pdf w

where; Ʃpdfw is defined as:

∑ pdf = ∑ pdf w ( l )
lmax
w l
(9)

At the end, the gamma (γ) value is obtained in the following way:
γ = 1− cdf w ( l ) (10)

1.3.3 MORPHOLOGICAL OPERATION-BASED ENHANCEMENT

Morphological operations make use of structuring elements while enhancing

retinal artifacts. Disk shaped masks are generally used among researchers in
biomedical image data enhancement. Tophat and bottom-hat transformations
are represented (Digital Image Processing, Gonzalez) in Eqns. (11) and (12),
respectively:
Tophat ( I=
) ( I − ( I ( open ) b ) ) (11)
Bottomhat
= ( I ) ( ( I ( close ) b ) − I ) (12)

where; b is structuring element and morphological operation (I(open)b)is

erosion of I by b, followed by dilation with b and (I(close)b is dilation of
I by b, followed by erosion with b. According to Hassanpour et al. (2015),
8 Advances in Data Science and Computing Technology

subsequent applications of top-hat and bottom-hat transformation lead to

contrast enhancement between light and dark arenas.
I + Tophat ( I ) − Bottomhat ( I )
En = (13)
When Eqn. (13) is tried with different structural elements of suitable size
enhancement result varies. The output enhanced retinal image is (En).

1.4 EVALUATION METRICS

We used two types of evaluation metrics for the assessment of retinal image
enhancement techniques. They are IQA metrics and classification quality
assessment (CQA) metrics.

1.4.1 IQA METRICS

IQA metrics are categorized in two classes: statistical error based metrics and
visual information based metrics. Ideally, a human observer only can provide
most consistent IQA metric to assess an image enhancement performance. It
is qualitative evaluation. Visual information may not be properly represented
by statistical error based metrics always instead human observer perception
can be quite aptly captured by visual information based metrics. PSNR and
AMBE are the mostly used statistical error based IQA metric and visual
information based IQA metric, respectively.

1.4.1.1 PEAK SIGNAL TO NOISE RATIO (PSNR)

PSNR is defined as:

( lmax )
2

PSNR = 10log (14)

MSE
MSE is mean squared error (MSE) between the original image(I) and the
enhanced image(En) and is defined as:
1
∑ ( En ( m,n ) − I ( m,n ) )
2
=MSE (15)
MN m,n

Higher PSNR value specifies more information content.

Deep Learning Supported Evaluation of Retinal Vessel 9

1.4.1.2 ABSOLUTE MEAN BRIGHTNESS ERROR (AMBE)

AMBE is the absolute difference between enhanced image expectation

(E(En)) and original image expectation (E(I)).
AMBE
= ( E ( En ) − E ( I ) ) (16)
Lower AMBE value specifies better image enhancement.

1.4.2 DEEP CLASSIFICATION METRICS

The significances of true positive (TP), false positive (FP), true negative
(TN), and false negative (FN) with respect to vessel classification are as
follows, TP is the proportion of actual vessel pixels in the whole popula­
tion whereas, TN is the proportion of actual non-vessel pixels in the whole
population. Similarly, FP is the proportion of falsely identified vessel pixels
in the whole population whereas FN is the proportion of falsely recognized
non-vessel pixels in the whole population. Accuracy and sensitivity are two
important evaluation metrics in case of medical image processing.

1.4.2.1 SENSITIVITY

Sensitivity (Sn) is TP rate. The closer the value of sensitivity to 1, the better
is the identification power of vessel pixel of the method.
=Sn TP / (TP + FN ) (17)

1.4.2.2 ACCURACY

Accuracy (Acc) provides proportion of both TP and true negative.

Acc = (TP + TN ) / (TP + TN + FP + FN ) (18)

1.5 DATASET DESCRIPTION

DRIVE (Digital retinal image for vessel extraction, Image Sciences Insti­
tute) retinal image vessel segmentation dataset is used for method evalu­
ation as it is the most frequently used dataset among researchers. Canon
10 Advances in Data Science and Computing Technology

CR5 non-mydriatic 3CCD camera with 45° field of view has been used
to capture retinal images. Each image was preserved with 8 bits per color
plane and at 565 x 584 pixels’ resolution. In this dataset ground truths are
also available.

1.5.1 PATHOLOGICAL SAMPLES

Around 40 photographs have been randomly considered, 33 are healthy

and 7 show symptoms of mild early diabetic retinopathy. Among all those
7 images 3 belongs to training set whereas 4 belongs to test set. A brief
description of the abnormalities is as available online (https://drive.grand­
challenge.org/):
• 25 Training: Pigment epithelium changes, probably buttery macu­
lopathy with pigmented scar in fovea, or choroidopathy, no diabetic
retinopathy or other vascular abnormalities.
• 26 Training: Background diabetic retinopathy, pigmentary epithelial
atrophy, atrophy around optic disk.
• 32 Training: Background diabetic retinopathy.
• 03 Test: Background diabetic retinopathy.
• 08 Test: Pigment epithelium changes, pigmented scar in fovea, or
choroidopathy, no diabetic retinopathy or other vascular abnormalities.
• 14 Test: Background diabetic retinopathy.
• 17 Test: Background diabetic retinopathy.

1.6 DEEP NEURAL NETWORK (DNN)

A deep CNN model has been implemented and then tested with differently
pre-processed image datasets. CNN architecture has been formed using U-net
type connections. U-net is a CNN, efficient in the field of medical image
segmentation. The characteristic of U-net is that it utilizes the same feature
maps for contraction of the image to a feature vector and for expansion of
the vector to a reconstructed image. Thus, structural integrity of image data
is retained. U-net model is trained on DRIVE trainset for vessel identifica­
tion. Application of U-net model on appropriately enhanced image dataset
leads to pixel level classification of vessel and non-vessel. Classification
efficiency measurement of the same model with differently enhanced image
dataset indicates better image enhancement method.
Deep Learning Supported Evaluation of Retinal Vessel 11

1.6.1 ARCHITECTURE

U-net architecture Ronneberger et al. (2015) is used here for CNN based
input IQA. Cross-entropy loss function and stochastic gradient descent
(SGD) optimization algorithm is used in implementation. Rectified linear
unit (ReLU) is used as activation function. A drop out of 0.2 has been
employed between two consecutive convolutional layers. The summary of
the train model is represented in Figure 1.1.

FIGURE 1.1 Summary of the deep CNN model.

12 Advances in Data Science and Computing Technology

1.6.2 TRAINING

DRIVE images undergo three types of image enhancement techniques

to obtain three different datasets as discussed in Section 1.3. Training is
performed on image patches of differently preprocessed image datasets.
Data augmentation has not been performed. During deep classification based
evaluation, 48 x 48 random overlapping patches are generated from image
dataset for proper training with less number of image samples. 90% of patch
dataset has been used for training whereas 10% of patch dataset has been
used for validation. Mini batch size is 32 patches.

1.6.3 IMPLEMENTATION REQUIREMENTS

All the experiments have been implemented in Python 3.6.8. High level APIs
of Google’s deep learning library TensorFlow have been used to implement
and evaluate the work. Implementations have been executed in Google
online Colab cloud environment where Tesla K80 GPU, 12 GB VRAM, 4*
Intel(R) Xeon(R) CPU @ 2.20 GHz, 13.51 GB RAM, 358.27 GB HDD are
available as deep network execution resources.

1.7 EXPERIMENTAL RESULTS

1.7.1 IMAGE QUALITY ASSESSMENT (IQA) METRIC BASED

EVALUATION

During enhancement, green channel of retinal image is only considered

(Nandy and Banerjee, 2012) as the intensity in green channel facilitates
retinal structures to be observable. Retinal images are almost saturated in
the red channel and have very low contrast in blue channel. Green channel
images are contrast enhanced by applying CLAHE. Another effective image
enhancement tool is Gamma correction. Green channel images undergo
adaptive linear gamma correction where all the pixels are linearly weighted
by some predefined factor (Huang et al., 2013). One more exhaustively used
retinal image enhancement tool is application of morphological operators
(Hassanpour et al., 2015). All four images and their histograms are given in
Figure 1.2. Average PSNR and AMBE for DRIVE images are shown both
graphically and in tabular form in Figure 1.3 and Table 1.1, respectively.
Deep Learning Supported Evaluation of Retinal Vessel 13

FIGURE 1.2 (a) Retinal green channel and histogram; (b) CLAHE and stretched histogram;
(c) morphology and stretched histogram; (d) AGC and stretched histogram.

FIGURE 1.3 Average PSNR and AMBE graph for DRIVE images.

TABLE 1.1 DRIVE Test Set Images – Average IQA Metrics Value

Enhancement Avg Image Quality Assessment Metrics Value
Techniques PSNR AMBE
Morphology 24.8719 5.7424
CLAHE 14.3472 35.6176
AGC 12.6987 47.6397

1.7.2 DEEP CLASSIFICATION METRIC BASED EVALUATION

48 x 48 random and overlapping image patches are generated from image

dataset. U-net type deep CNN architecture has been fed with patches for
14 Advances in Data Science and Computing Technology

training. A total of 1,71,000 training patches and 19,000 validation patches

were generated from DRIVE train set images. The model was trained for 10
epochs only, as the trend of classification result can be observed from less
number of epochs also. Accuracy and sensitivity are two important evalua­
tion metrics in the field of vessel classification. Average training accuracy,
test accuracy and sensitivity obtained from differently enhanced image
datasets are in Table 1.2. Graphs for average training and test accuracies
and receiver operator characteristic graph of corresponding classification are
also presented from Figure 1.4 to Figure 1.6.

TABLE 1.2 DRIVE Test Set Images – Average Deep Classification Metrics
Enhancement Avg Classification Quality Assessment Metrics Value
Techniques Train_Acc Test_Acc Sensitivity
Morphology 0.9576 0.9524 0.7573
CLAHE 0.9568 0.9525 0.7564
AGC 0.9510 0.9484 0.7349

FIGURE 1.4 ROC and accuracy curves of morphologically enhanced images.

FIGURE 1.5 ROC and accuracy curves of CLAHE images.

Deep Learning Supported Evaluation of Retinal Vessel 15

FIGURE 1.6 ROC and accuracy curves of AGC enhanced images.

1.7.2 IMAGE QUALITY ASSESSMENT (IQA) OF PATHOLOGY

SAMPLES

Four images among 20 drive test images have mild early DR signs as patho­
logical symptoms. Table 1.3 shows individual performance measurements
obtained by applying deep CNN classification on those pathology samples
separately. This is a reverse view point to assess the applicability of different
enhancement techniques on abnormal images also. In most of the cases
morphologically enhanced abnormal images are characterized with higher
metric values.

TABLE 1.3 Deep CNN Classification Metrics on DRIVE Test Dataset Abnormal Images
Symptoms Enhancement Deep Classification Assessment Metrics
Techniques Accuracy Sensitivity
Hard exudate Morphology 0.9420 0.6921
CLAHE 0.9420 0.6928
AGC 0.9420 0.7670
Hard exudate Morphology 0.9464 0.6580
CLAHE 0.9458 0.6489
AGC 0.9365 0.633
Hemorrhage Morphology 0.9574 0.8164
CLAHE 0.9571 0.8106
AGC 0.9535 0.7987
hard exudate Morphology 0.9493 0.6864
CLAHE 0.9515 0.7153
AGC 0.9487 0.7082
16 Advances in Data Science and Computing Technology

1.8 DISCUSSION

Statistical error based and visual perception based IQA metrics are
considered in this work for preliminary image enhancement evaluation.
Here, PSNR represents statistical error based metrics. But it is much more
important to verify evaluation with human visual perception based metric
such as AMBE. Maximum PSNR value specifies high information content
of the morphologically enhanced image. Again, minimum AMBE value of
morphologically enhanced image also identifies it as better enhanced image.
Table 1.1 data represents the fact properly. Average values are computed on
DRIVE test dataset images.
This observation is further estimated from the view point of deep CNN
classification as vessel enhancement is a very important prior sub part of
retinal vessel segmentation. So, better enhanced images tend to produce better
classification efficiency metrics. Maximum value of average segmentation
accuracy indicates the superior quality of morphologically enhanced image data
with application to retinal fundus images. In the field of vessel segmentation
use of sensitivity as performance measurement index is highly significant.
Maximum average sensitivity has also been obtained from morphologically
enhanced test images. Table 1.2 represents average values of train accuracy,
test accuracy and sensitivity on DRIVE images. Different ailment symptoms
are evident in retina at a very early phase. Table 1.3 provides individual test
accuracy and sensitivity of pathology samples of DRIVE test set separately.
In individual evaluation of pathology samples also, most of the classification
efficiency values are in favor of morphological operation.
Though morphological enhancement performed very well in both IQA
based evaluation as well as deep classification based evaluation, performing
image enhancement by combining enhancement methods in some specific
way may produce better result as well.

1.9 CONCLUSION

Three single image enhancement techniques are experimented in this

work for their efficiency in image enhancement with application to retinal
vessel segmentation. These single enhancement techniques are applica­
tions of CLAHE, AGC, and morphological enhancements. From the view
point of IQA metric values it is decided quantitatively that morphological
enhancement performs well for this purpose. This claim has been further
established by deep CNN vessel classification based evaluation also as better
Deep Learning Supported Evaluation of Retinal Vessel 17

classification metrics are expected to obtain from better vessel enhanced

input image datasets.
In vessel segmentation literature, these prevalently used enhancement
techniques are used in different arbitrary combinations also. It is intended to
enrich this work by analyzing combinations of these techniques in different
possible orders and evaluating which one is supposed to yield even better
results. Deep learning based result verification may be extended to cross
dataset evaluation, too. In this way, the best combination of enhancement
techniques relevant to the pertinent field can be decided. This may be consid­
ered as the future scope of this work.

KEYWORDS

• adaptive gamma correction (AGC)

• contrast limited adaptive histogram equalization (CLAHE)
• convolutional neural network (CNN)
• image quality assessment (IQA)
• morphology

REFERENCES

Almotiri, J., Elleithy, K., & Elleithy, A., (2018). Retinal vessels segmentation techniques and
algorithms: A survey. Applied Sciences, 8.
Cai, J., Gu, S., & Zhang, L., (2018). Learning a deep single image contrast enhancer from
multi-exposer images. IEEE Transactions on Image Processing, 27, 2049–2062.
Fu, H., Xu, Y., Lin, S., Wong, D., & Liu, J., (2016). Deep vessel: Retinal vessel segmentation
via deep learning and conditional random field. Proceedings of Medical Image Computing
and Computer Assisted Intervention MICCAI, 132–139.
Gonzalez, R. C., & Woods, R. E. (2015). Digital Image Processing (3rd edn.). Pearson.
Hassan, G., El-Bendary, N., Hassanien, A., Fahmy, A., Shoeb, A., & Snasel, V., (2015).
Retinal blood vessel segmentation approach based on mathematical morphology. Procedia
Computer Science, 65, 612–622.
Hassanpour, H., Samadiani, N., & Mahdi, S. S. M., (2015). Using morphological transforms
to enhance the contrast of medical images. The Egyptian Journal of Radiology and Nuclear
Medicine, 46, 481–489.
Huang, S., Cheng, F., & Chiu, Y., (2013). Efficient contrast enhancement using adaptive
gamma correction with weighting distribution. IEEE Transaction on Image Processing, 22.
18 Advances in Data Science and Computing Technology

Liskowski, P., & Krawiec, K., (2016). Segmenting retinal blood vessels with deep neural
networks. IEEE Transactions on Medical Imaging, 35, 2369–2380.
Mo, J., & Zhang, L., (2017). Multi-level deep supervised networks for retinal vessel
segmentation. International Journal of Computer Assisted Radiology and Surgery, 12,
2181–2193.
Nandy, M., & Banerjee, M., (2012). Retinal vessel segmentation using Gabor filter and
artificial neural network. Proceedings of IEEE International Conference on Emerging
Applications of Information Technology – EAIT, 157–160.
Nandy, M., & Banerjee, M., (2019). Automatic diagnosis of micro aneurysm manifested
retinal images by deep learning method. Proceedings of ICETST.
Nandy, P. M., & Banerjee, M., (2017). A comparative analysis of application of niblack and
sauvola binarization to retinal vessel segmentation. Proceedings of CINE. IEEE Xplore.
Pizer, S., Amburn, E., Austin, J., Cromartie, R., Geselowitz, A., Greer, T., Ter, H. R. B., et al.,
(1987). Adaptive histogram equalization and its variations. Computer Vision, Graphics and
Image Processing, 39, 355–368.
Rafael C. Gonzalez & Richard E. Woods, (2008). Digital Retinal Images for Vessel Extraction
(DRIVE) Dataset. Image Sciences Institute, http://www.isi.uu.nl/Research/Databases/
DRIVE/ (accessed on 8 December 2021).
Rahman, S., Rahman, M., Abdullah-Al-Wadud, M., Al-Quaderi, G., & Shoyaib, M., (2016).
An adaptive gamma correction for image enhancement. EURASIP Journal on Image and
Video Processing, 35, 1–13.
Ronneberger, O., Fischer, P., & Brox, T., (2015). U-Net: Convolutional Networks for
Biomedical Image Segmentation. arXiv:1505.04597v1 [cs.CV].
Roychowdhury, S., Koozekanani, D., & Parhi, K. K., (2015). Iterative vessel segmentation of
fundus images. IEEE Transactions on Biomedical Engineering, 62, 1738–1749.
Sheba, K., & Gladston, R., (2016). Objective quality assessment of image enhancement
methods in digital mammography-a comparative study. Signal and Image Processing: An
International Journal, 7.
Singh, S., & Bovis, K., (2005). An evaluation of contrast enhancement techniques for
mammographic breast masses; Chapter VIII. 5. Graphics Gems, IV., 9, 109–119.
Staal, J., Abrano, A., Niemeijer, M., Viergever, M., & Ginneken, B., (2004). Ridge based
vessel segmentation in color images of the retina. IEEE Transactions on Medical Imaging,
23, 501–509.
Zuiderveld, K., (1994). Contrast limited adaptive histogram equalization. Chapter VIII.5,
Graphics Gems, IV, 474–485.
CHAPTER 2

AN EFFICIENT AND OPTIMIZED

DEEP NEURAL NETWORK
ARCHITECTURE FOR HANDWRITTEN
DIGIT RECOGNITION
SHUVANKAR ROY and MAHUA NANDY PAL
Computer Science and Engineering Department,

MCKV Institute of Engineering, Howrah–711204, West Bengal, India,

E-mail: [email protected] (S. Roy)

ABSTRACT

This work presents a new, application oriented and hyper parameter

optimized architecture of deep neural network (DNN) for hand written digit
recognition. Though a substantial amount of work has been done in this
field, yet it is an open area of research as there is scope to determine the
suitable deep architecture for a particular problem domain and to improve the
efficiency of deep classifier through optimization. Evaluating the efficiency
of deep classification process and further improvement of its performance
in the relevant field leads to the development of a competent network. In
this chapter, A DNN architecture, has been developed which is capable of
recognizing hand written digits in a very efficient way. Classification accuracy
is used as evaluation metric. The proposed DNN architecture is capable of
achieving almost 96.5% accuracy before hyper parameter optimization and
shows a further improved accuracy value of 98.12% after hyper parameter
optimization. These results are indicative of its relevance in the particular
domain.
20 Advances in Data Science and Computing Technology

2.1 INTRODUCTION

Before establishment of proper technologies, we had been heavily relied on

hand written materials and data, which is error prone and even difficult to store
and access. Due to the wave of digitization, the previously necessary hand
written documents are needed to be digitized. Handwritten digit recognition is
a subpart of the same problem. Digitized handwritten documents can be made
usable by adopting a complete highly accurate handwritten text recognition
tool. Hand written character and digit recognition make the whole process of
storage, organization, and access of hand written data comparatively simpler.
Adoption of a highly accurate handwritten text recognition tool makes user
authentication problem easier.
As the system has immense practical importance, deep learning has
been applied to it to improve its efficiency. Currently deep learning applica­
tion has become state-of-the-art for various classification problem for its
outstanding performance in the relevant field. But not all network architec­
tures are suitable for various problem domains. So, deciding proper network
architecture with optimized hyper parameters is very important with respect
to a particular application domain. Though training of deep neural network
(DNN) is computationally expensive and very time consuming, optimization
of hyper parameters is expected to yield better network performance.

2.2 LITERATURE SURVEY

Now-a-days deep learning exploration has become the recent trend of research
in various fields. With advent of high-end architecture suitable to implement
deep learning environment, researchers try to exploit the benefit of this to a
greater extent. Among different handwritten digit or character classification
methods, Hochreiter and Schmidhuber (1997), and Voigtlaender, Doetsch,
and Ney (2016) describe applications of long short-term memory (LSTM).
Graves, Fernandez, Gomez, and Schmidhuber (2006) and Graves et al.
(2009) trained recurrent neural network (RNN) with connectionist temporal
classification (CTC). Graves and Schmidhuber (2009) also used a multidi­
mensional RNN. Pham, Bluche, Kermorvant, and Louradour (2014) adjusted
neural network parameters to improve the result. Bluche and Messina (2017)
deals with multi lingual hand writing recognition. Puigcerver (2017) verifies
the applicability of different recurrent layers whereas Castro, Bezerra, and
Valena (2018) establishes the effectiveness of boosting multidimensional
An Efficient and Optimized Deep Neural Network 21

LSTM for hand written recognition system. DNNs outperform on many

machine learning problems, but they are very sensitive to hyperparameters
setting (Domhan, Springenberg, and Hutter, 2015).
Digit recognition system has been evaluated using different automated
learning algorithms namely, Multilayer Perceptron, support vector machine
(SVM), Naïve-Bayes, Bayes Net, Random Forest, J48 and Random Tree
using WEKA in 2018 (Shamim, Miah, Sarker, Rana, and Jobair, 2018). The
chapter (Zohra and Rao, 2019) presented performance analysis of SVM,
K-nearest neighbor (KNN) and convolutional neural network (CNN).
Deep learning is a powerful technique in current generation. The chapter
(Mishra, Malathi, and Senthilkumar, 2018) dealt with handwritten digit recog­
nition on the well-known image database using Convolution neural network.
Deep learning increases accuracy. In the process, the model becomes highly
efficient to predict distorted data also. Additionally, a comparative analysis
of the model has been performed on multicore CPU and General GPU and
computation times are recorded. In this chapter (Vijayalaxmi, Rudraswami­
math, and Bhavanishankar, 2019), the Handwritten Digit Recognition using
Deep learning methods has been implemented. The maximum used machine
learning algorithms were compared by training and testing on same dataset.
The classifiers used are KNN, SVM, RFC, and CNN. Utilizing these deep
learning techniques, a high amount of accuracy can be obtained. The goal of
the chapter (Siddique, Sakib, and Siddique, 2019) is to analyze the effective
number of hidden layers and epochs and corresponding accuracies. They
performed experiments on Modified National Institute of Standards and
Technology (MNIST) dataset to evaluate their work. Lastly, Baldominos,
Saez, and Isasi, (2019) has provided an exhaustive review of the state of the
art for both the MNIST and EMNIST databases.
In industries and academic fields, the demand for the type written and hand
written character or digit recognition is always there. Character recognition
is a combination of detecting, segmenting, and identifying characters from an
image. The ultimate goal of hand written character recognition is to simulate
the human reading capabilities so that machine can read, understand, edit,
and work with text similar to human. Handwritten character recognition is
more difficult than type written one. Type written texts are equally spaced and
have a specified size of each character, which makes character segmentation
a lot easier. Whereas handwritten texts may not be accurately spaced, may
contain skewness and non-horizontal line, malformed font style and even
non-human understandable characters, which makes the process of character
extraction a tougher job. A big challenge in multi-lingual hand written text
22 Advances in Data Science and Computing Technology

recognition is non-availability of large corpus of hand written data. The

results of hand written digit or character recognition using deep learning
produces better results compared to the same using traditional machine
learning in most of the time.
The main focus of this chapter is to provide a framework for deep learning
architecture in the field of hand written digit recognition. The current litera­
ture indicates that the challenges present in the field of deep learning are
proper application-oriented network architecture, parameter optimization,
minimization of time complexity and computational complexity, require­
ment of huge sample data for network learning, etc. So, determining proper
application specific deep architecture and its hyper parameter optimization
for hand written digit recognition is very significant in this respect.
Following sections represent theoretical background, implementation,
and results, discussion, and conclusion of the chapter, respectively.

2.3 THEORETICAL BACKGROUND

2.3.1 CLASSIFIER OPTIMIZATION

Classifier optimization problem deeply relies on formulating four factors.

They are identification of parameter search space, minimization or
maximization of objective function, the basic model on which search space
values will be evaluated and stored outcomes of each evaluation. In case of
DNN optimization, different network hyper parameters that are intended to
optimize in this chapter are learning rate, learning algorithm used, number
of hidden layers, batch size and dropout. These terms are explained in the
following subsections. The search space for each individual parameter is
defined. The objective functions used here are validation accuracy and test
accuracy.

2.3.2 DEEP NEURAL NETWORK (DNN)

A DNN is an artificial neural network having more than one hidden layer in
between input and output layer. The neurons of the network are inter connected
with weighted connections. The advantage of using deep network is that the
network itself is capable of extracting features of the input patterns. Thus,
programmer does not require to establish efficient handpicked features to repre­
sent input patterns. The DNN computes and adjusts the correct mathematical
An Efficient and Optimized Deep Neural Network 23

weights to convert the input into the output. The impulse or signal passes
through the neural network layers finding the probability of each output.

2.3.3 NETWORK LAYERS

Neural network has three types of layers made up of a number of neurons.

They are the input layer, hidden layer, and output layer.

2.3.3.1 INPUT LAYER

Input layer accepts initial information from outside the network. The consecu­
tive nodes of the network process these inputs to draw some inference.

2.3.3.2 HIDDEN LAYER

These are the layers in between input layer and output layer of a neural
network. The nodes of theses layers are responsible for processing the inputs
and all intermediate computations.

2.3.3.3 OUTPUT LAYER

Output layer is responsible for producing the output of the neural network by
computing individual class probability.

2.3.4 NETWORK HYPER PARAMETERS

Network hyper parameters are some values which are required to be decided
before starting training process of the network such as learning algorithm,
leaning rate, number of hidden layers, batch size, dropout value, etc. Deciding
the best evaluated value of hyper parameters is known as hyper parameter
tuning. Hyper parameter tuning may be done either manually or by applying
grid search over the search space.
Following are some of the network hyper parameters which have been
tried to be optimized or tuned in the following subsections for network
performance improvement.
24 Advances in Data Science and Computing Technology

2.3.4.1 LEARNING RATE

In machine learning and statistics, the learning rate is a tuning parameter

that defines the step size to be taken at each iteration of learning algorithm
while moving towards a minimum of loss function. The lower the value of
learning rate, the slower we travel towards minimum and the lower is the
chance of missing local minima, but low learning rate also means the model
learning phase will be time consuming and will take long time to converge if
the model is stuck in plateau region.

2.3.4.2 OPTIMIZER

Optimizer or optimization algorithm can be used interchangeably. Optimiza­

tion algorithm determines the speed and efficiency of a machine learning
algorithm. Different optimization algorithms are stochastic gradient descent
(SGD), mini-batch gradient descent, gradient descent with momentum,
Adam optimizer, etc.
SGD method iteratively calculates stochastic approximation of gradient
descent optimization. SGD replaces the actual gradient value, calculated
from the entire data set by an estimated value, calculated from a randomly
selected subset of data.
Adam optimization algorithm can be considered as a combination of
RMSprop and SGD with momentum. It utilizes squared gradients to scale
up the learning rate like RMSprop and uses of momentum by using moving
average value of the gradient instead of gradient itself like SGD with
momentum.
Adagrad is a learning algorithm with parameter-specific learning rates.
The algorithm depends on frequency of updating of a parameter during
training. Learning rate value is inversely proportional to the update frequency
that is the greater number of a parameter receives update, the smaller is the
value of learning rate.

2.3.4.3 HIDDEN LAYERS

Features of input layers are propagated through hidden layers. Number of

hidden layers should also be determined for obtaining proper output.
An Efficient and Optimized Deep Neural Network 25

2.3.4.4 BATCH SIZE

It indicates number of training samples presented in one iteration. The entire

dataset is divided into number of small batches. Ideally the batch size may
be within the range from 1 to the whole training sample set.

2.3.4.5 DROPOUT

Dropout is a regularization method and is used to prevent overfitting during

neural network training. By drop out some neurons are randomly removed
from each layer and not used in forward or back propagation. The signifi­
cance of the dropout hyperparameter is the probability of training a given
node in a layer, where 1.0 means no dropout, and 0.0 means no outputs from
the layer.

2.4 IMPLEMENTATION AND RESULTS

2.4.1 DATASET DESCRIPTION

MNIST (modified national institute of standards and technology database)

is a large collection of handwritten digits that is commonly used for training
and testing in the fields of various image processing systems as well as
machine learning systems. The images in the MNIST database are from two
NIST’s databases: Special Database 1 and Special Database 3. Special Data­
base 1 and Special Database 3 digits were written by high school students
and employees of the United States Census Bureau, respectively. The dataset
contains 60,000 training sample images and 10,000 test sample images of
28 pixels in height and 28 pixels in width, for a total of 784 pixels in total.
Each pixel is associated with an intensity value between 0 and 255 inclusive,
indicating lightness or darkness of that pixel. The original black and white
(binary) images were size normalized to 20×20 pixel while preserving the
aspect ratio. The normalization algorithm uses anti-aliasing technique and
thus the resulting images contain gray levels. The images were centered in a
28×28 image by computing the center of mass of the pixels, and translating the
image so as to position this point at the center of the 28×28 field. Figure 2.1
represents some examples of handwritten digits of MNIST dataset.
26 Advances in Data Science and Computing Technology

FIGURE 2.1 Example of handwritten digits in MNIST database.

Each row of the csv file represents one labeled example. As this is a
10-class classification problem, column 0 represents the label that a human
observer has assigned for it. Columns 1 through 784 represent the feature
values for each pixel of 28×28=784-pixel values. The pixel values are in
the range 0–255. 0 represents white, 255 represents black. Most of the pixel
values are 0. Numbers of training samples and test samples are 48,000 and
12,000, respectively. The ratio of train set and validation set is 80:20.

2.4.2 EVALUATION METRICS

Validation accuracy and test accuracy have been used as evaluation metrics
in this process. Validation set is a part of training set, used to validate the
model during its generation. Validation phase avoids overfitting of the
model. Test set is the unknown set of data, which the already built model is
supposed to classify.
True positive (TP) is the proportion of truly positive samples in the whole
population whereas, true negative (TN) is the proportion of truly negative
samples in the whole population. Similarly, false positive (FP) is the propor­
tion of falsely positive samples in the whole population whereas false negative
(FN) is the proportion of falsely negative samples in the whole population.
Accuracy can be defined as the proportion of true results, both TP and
TN, in a population. It can be defined in terms of TP, TN, FP and FN as:
Accuracy = (TN + TP)/(TN + TP + FN + FP) (1)
An Efficient and Optimized Deep Neural Network 27

2.4.3 IMPLEMENTATION REQUIREMENTS

All the experiments have been implemented in Python 3.6.8 using Tensor­
Flowv1 at backend. Implementations have been executed in Google online
Colab cloud environment where Tesla K80 GPU, 12 GB VRAM, 4* Intel(R)
Xeon(R) CPU @ 2.20 GHz, 13.51 GB RAM, 358.27 GB HDD are available
as deep network execution resources.
High level APIs of Google’s deep learning library TensorFlow have been
used to implement and evaluate the work.
The input layer contains 784 weights corresponding to the input images
shape of 28×28 pixel. The first hidden layer will contain 784×n weights where
n represents the number of nodes in the layer. First hidden layer captures
low level features and consecutive layers capture more sophisticated features
which can be visualized for perception also.

2.4.4 NETWORK ARCHITECTURE

The network architecture followed to evaluate the work consists of one input,
one output layer and 4 hidden layers. Rectified linear unit (ReLU) has been
used as nonlinear activation function in hidden layers. One dropout layer has
been inserted between 4th hidden layer and output layer with dropout value
of 0.2. Different layers, their output shape and number of parameters in each
layer have been depicted in Figure 2.2.

2.4.5 HYPER PARAMETER OPTIMIZATION

2.4.5.1 LEARNING RATE EVALUATION

Performance of the model has been evaluated using three different learning
rates and the results obtained are represented in Table 2.1. Best validation
accuracy and test accuracy are secured using learning rate = 0.05. Table 2.1
represents learning rate evaluation.

2.4.5.2 OPTIMIZATION ALGORITHM EVALUATION

Performances of using three optimization algorithms during training model

generation have been evaluated keeping best performing learning rate = 0.05
28 Advances in Data Science and Computing Technology

as constant. These optimization algorithms are Adagrad, Adam, and Gradient

Descent. Table 2.2 represents optimization algorithm evaluation.

FIGURE 2.2 Summary of the model.

TABLE 2.1 Learning Rate Evaluation

Learning Rate Validation Accuracy Test Accuracy

0.5 0.61 0.61
0.1 0.95 0.95
0.05 0.95 0.96
0.01 0.93 0.93

TABLE 2.2 Optimization Algorithm Evaluation

Optimizer Validation Accuracy Test Accuracy
Adagrad optimizer 0.95 0.96
Adam optimizer 0.86 0.87
Gradient descent optimizer 0.94 0.94

2.4.5.3 NUMBER OF HIDDEN LAYERS

Network performance has been evaluated by taking 1, 2, 3, 4, and 5 number

of hidden layers while previously discussed parameters are kept fixed at
An Efficient and Optimized Deep Neural Network 29

their best performing values. Table 2.3 represents number of hidden layer
evaluation.

TABLE 2.3 Optimal Number of Hidden Layer Evaluation

Number of Hidden Layers Validation Accuracy Test Accuracy
1 0.9792 0.9788
2 0.9828 0.9815
3 0.9823 0.9797
4 0.9835 0.9812
5 0.9818 0.9791

2.4.5.4 BATCH SIZE EVALUATION

Batch size of 30, 60, 100, and 150 have been experimented for network
performance evaluation. Table 2.4 shows all the metric values. Best perfor­
mance is provided if batch size considered is 60.

TABLE 2.4 Batch Size Evaluation

Batch Size Validation Accuracy Test Accuracy
30 0.9834 0.9811
60 0.9837 0.9821
100 0.9835 0.9812
150 0.9827 0.9793

2.4.5.5 DROPOUT EVALUATION

Performances of the model has been evaluated by introducing a dropout

layer after hidden layers keeping best performing learning rate = 0.05,
number of hidden layers = 4, batch size = 60 and Adagrad as best
performing optimization algorithm. Best Accuracy has been achieved with
dropout value of 0.2 beyond which performance of the network decreases
drastically.
Test accuracy value of 0.98 using dropout is more significant than
the same accuracy obtained without dropout as drop out decreases time
and computational complexity of deep learning training phase to a great
extent.
30 Advances in Data Science and Computing Technology

2.4.6 GRAPHICAL REPRESENTATION OF HYPER PARAMETER

EVALUATION

LogLoss curves are plotted to represent the performances of the classifier

model. Figures 2.3(a–c) represent the same for three different optimization
algorithms. They are Adagrad, Adam, and SGD respectively. During evalua­
tion of optimization algorithm performances learning rate is set to the value
of 0.05. Figure 2.3(d) represents the application of regularization method.
The LogLoss curves for dropout introduction with a value of 0.2 is also
very interesting as the same accuracy have been obtained after introduc­
tion of dropout too. These plots have been obtained while keeping learning
rate, number of hidden layer and batch size fixed to the optimized value and
taking Adagrad as optimization algorithm.

(a) (b)

(c) (d)
FIGURE 2.3 LogLoss curves (a) Adagrad; (b) Adam; (c) SGD; (d) Adagrad with lr = 0.05
and dropout = 0.3.

Finally, the curves for training/validation accuracy and training/validation

loss with optimized classifier have been shown separately in Figures 2.4(a)
and (b), respectively.
An Efficient and Optimized Deep Neural Network 31

(a) (b)
FIGURE 2.4 (a) Training and validation accuracy of optimized network; (b) training and
validation loss of optimized network.

2.5 DISCUSSION

Among different DNN hyper parameters, learning rate (lr), learning algorithm,
number of hidden layers, dropout value and batch size have been addressed
to frame the optimization problem. Optimizations of these hyper parameters
are evaluated by maximization of objective function. The objective functions
used here are test and validation accuracies. The stored outcome of best
evaluation indicates optimized hyper parameter values.
The search space of the optimization problem has been defined as follows.
Four lr values (0.5, 0.1, 0.05, and 0.01) have been evaluated. Among them
best accuracy was obtained with lr = 0.05. Different learning algorithms such
as SGD, Adam, and Adagrad have been evaluated with lr = 0.05. Adagrad
algorithm performed best to the specific application of hand written digit
recognition.
This classifier has been further optimized for number of hidden layers. Five
variations (1, 2, 3, 4, and 5 hidden layers) have been considered for evaluation.
Following the experimental values obtained for objective functions, it can be
decided that the network architectures with two or four hidden layers perform
the best. Moreover, the classifier has been experimented with variations of
batch sizes (30, 60, 100, and 150). The best performing batch size is 60.
With these four network hyper parameter optimizations, the model
is further optimized in terms of time and computational complexity by
applying regularization method, i.e., introducing dropout layers in between
last hidden layer and output layer as dropout layer randomly drops out some
definite percentage of hidden neurons from the network. In this process, the
same accuracy has been obtained with 0.2 dropout value also. Maximum
32 Advances in Data Science and Computing Technology

permissible dropout value is 0.2 as per the network performance concerned.

Dropout value more than 0.2 degrades the network performance severely.
This should also be noted that the training has been conducted for only 20
epochs as the trend of classification can be perceived from a fewer number of
epochs also. Model training with a greater number of epochs, will be more
efficient to provide more accuracy value.

2.6 CONCLUSION

In this chapter, hyper parameter optimization problem of DNN environment

for hand written digit recognition has been addressed. It can be concluded
from this work that the Adagrad optimizer with learning rate of 0.05 may be
effective selections for the relevant field. Additionally, four hidden layers
and the batch size of 60 are efficient for the problem domain. Though accu­
racy has not improved with introduction of 0.2 dropout value, yet inclusion
of dropout reduces computational and time complexity of DNN.
The work may be enriched by extending it for multi lingual digit recogni­
tion. It may be a good future direction to automate and speed up network
hyper parameter optimization so that the process can cover up a large search
space in restricted time frame.

KEYWORDS

• batch size
• deep neural network
• dropout
• handwritten digit recognition
• hidden layer
• learning rate
• optimization algorithm

REFERENCES

Baldominos, A., Saez, Y., & Isasi, P., (2019). A survey of handwritten character recognition
with MNIST and EMNIST. Applied Sciences, 9(15), 3169. doi: 10.3390/app9153169.
An Efficient and Optimized Deep Neural Network 33

Bluche, T., & Messina, R., (2017). Gated convolutional recurrent neural networks for
multilingual handwriting recognition. In: 2017 14th IAPR International Conference on
Document Analysis and Recognition (ICDAR). doi: 10.1109/icdar.2017.111.
Castro, D., Bezerra, B. L. D., & Valenca, M., (2018). Boosting the deep multidimensional
long-short-term memory network for handwritten recognition systems. In: 2018 16th
International Conference on Frontiers in Handwriting Recognition (ICFHR). doi: 10.1109/
icfhr-2018.2018.00031.
Domhan, T., Springenberg, J. T., & Hutter, F., (2015). Speeding up automatic hyperparameter
optimization of deep neural networks by extrapolation of learning curves. Twenty-Fourth
International Joint Conference on Artificial Intelligence (IJCAI).
Graves, A., & Schmidhuber, J., (2009). Offline handwriting recognition with multidimensional
recurrent neural networks. Advances in Neural Information Processing Systems, 545–552.
Graves, A., Fernández, S., Gomez, F., & Schmidhuber, J., (2006). Connectionist temporal
classification. Proceedings of the 23rd International Conference on Machine Learning –
ICML 06. doi: 10.1145/1143844.1143891.
Graves, A., Liwicki, M., Fernandez, S., Bertolami, R., Bunke, H., & Schmidhuber, J., (2009).
A novel connectionist system for unconstrained handwriting recognition. IEEE Transactions
on Pattern Analysis and Machine Intelligence, 31(5), 855–868. doi: 10.1109/tpami.2008.137.
Hochreiter, S., & Schmidhuber, J., (1997). Long short-term memory. Neural Computation,
9(8), 1735–1780. doi: 10.1162/neco.1997.9.8.1735.
Mishra, S., Malathi, D., & Senthilkumar, K., (2018). Digit recognition using deep learning.
International Journal of Pure and Applied Mathematics, 118(22), 295–302. ISSN: 1314-3395.
Pham, V., Bluche, T., Kermorvant, C., & Louradour, J., (2014). Dropout improves recurrent
neural networks for handwriting recognition. In: 2014 14th International Conference on
Frontiers in Handwriting Recognition. doi: 10.1109/icfhr.2014.55.
Puigcerver, J., (2017). Are multidimensional recurrent layers really necessary for handwritten
text recognition? In: 2017 14th IAPR International Conference on Document Analysis and
Recognition (ICDAR). doi: 10.1109/icdar.2017.20.
Shamim, S. M., Miah, M. B. A., Sarker, A., Rana, M., & Jobair, A. A., (2018). Handwritten
digit recognition using machine learning algorithms. Indonesian Journal of Science and
Technology, 3(1), 29. doi: 10.17509/ijost.v3i1.10795.
Siddique, F., Sakib, S., & Siddique, M. A. B., (2019). Handwritten Digit Recognition Using
Convolutional Neural Network in Python with Tensor Flow and Observe the Variation of
Accuracies for Various Hidden Layers. doi: 10.20944/preprints201903.0039.v1.
Vijayalaxmi, R. R., & Bhavanishankar, K., (2019). Handwritten digit recognition using CNN.
International Journal of Innovative Science and Research Technology, 4(6). ISSN: 2456-2165.
Voigtlaender, P., Doetsch, P., & Ney, H., (2016). Handwriting recognition with large
multidimensional long short-term memory recurrent neural networks. In: 2016 15th Inter­
national Conference on Frontiers in Handwriting Recognition (ICFHR). doi: 10.1109/
icfhr.2016.0052.
Yann LeCun (1998). The MNIST Database. http://yann.lecun.com/exdb/mnist/ (accessed on
08 December 2021).
Zohra, M., & Rao, D. R., (2019). A comprehensive data analysis on handwritten digit recogni­
tion using machine learning approach. International Journal of Innovative Technology and
Exploring Engineering (IJITEE), 8(6). ISSN: 2278–3075.
CHAPTER 3

DESIGN AND DEVELOPMENT OF A

DRIVING ASSISTANCE AND SAFETY
SYSTEM USING DEEP LEARNING
SHEKHAR VERMA,1 TUSHAR BAJAJ,1 NIDHI SINDHWANI,1 and
ABHIJIT KUMAR2
Department of ECE, Amity School of Engineering and Technology,
1

New Delhi GGSIPU, Delhi, India

2
School of Computer Science, University of Petroleum and Energy
Studies, Dehradun, Uttarakhand, India,
E-mail: [email protected]

ABSTRACT

Road conditions are one of the major cause of accidents around the world.
Distress on road and combination of human ignorance/error may be really
fatal. This chapter proposes a system that can help to prevent road accidents
caused by distress on roads and human error. The proposed system minimizes
the occurrence of road accidents through real-time monitoring of road
conditions and the driver’s activities. A robust system has been proposed
based on image processing, computer vision and deep learning. This system
employs two cameras, one fixed in front of the vehicle for capturing the input
video frames which is being processed in real time using image processing
and computer vision algorithms, this detects the potholes or any sort of
distress on the roads which will be projected on the driver’s windshield using
a heads up display (HUD). Another camera will check the driver’s attention,
whether or not driver is operating radio or checking phone, talking over

Advances in Data Science and Computing Technology: Methodology and Applications. Suman Ghosal,
Amitava Choudhury, Vikram Kumar Saxena, Arindam Biswas, & Prasenjit Chatterjee (Eds.)
© 2023 Apple Academic Press, Inc. Co-published with CRC Press (Taylor & Francis)
36 Advances in Data Science and Computing Technology

mobile, etc. These two systems works along with a decision maker algorithm
which decides the severity of the situation and generate an alert signal to the
driver, if the road conditions are poor and if, the driver is distracted. A deep
learning based architecture applied to enhance the efficiency for monitoring
driver’s attention, which used StateFarm dataset.

3.1 INTRODUCTION

The number of vehicles operating in the world is increasing day by day.

There are millions of reported road accidents in the world in which thousands
of people lose their life every day. In a country like India, it was reported in
2019, that over 151,000 people died in road accidents (India Today, https://
www.indiatoday.in/india/story/over-9300-deaths-25000-injured-in-3-years­
due-to-potholes-1294147-2018-07-24; Guardian, https://www.theguardian.
com/world/2018/jul/25/more-deadly-than-terrorism-potholes- responsible­
for-killing-10-people-a-day-in-india). In times like this, the safety of
passengers is of utmost importance. The accidents might be caused by not
following the traffic rules, bad situations of the roads, weather, or human
errors. In another report, it was stated that over 30% of the accidents caused
in India were due to the poor road conditions, i.e., Potholes, Disintegration,
etc. Potholes are circular shape openings or depressions that can be caused by
water in the underlying soil surface and weathering of the road surface. They
appear when the top layer of the road, has been eroded by busy traffic thus
exposing the concrete base or structure underneath. Potholes are hazardous
for drivers and vehicles when moving at high speeds because it is very hard
for the driver to see the potholes on the road surface. Moreover, they possess
a threat to the car tires as they can rupture from the sudden impact leading
to accidents. Techniques, such as Machine Learning, Artificial Intelligence
and computer vision can be used in the vehicles to avoid and prevent these
situations. Vehicles should come with some sort of integrated infrastructure
in a way that this chapter suggests in order to reduce the chance so accidents
due to human error and road conditions. This system uses an approach of
determining the state of roads as well as the state of driver simultaneously
and producing alerts in real time which could even help in providing data
of the roads for improvement along with significant help in driving and
ensuring safety. The images captured by the road camera and the dashboard
camera will be processed using an onboard computer (for example, a SOC
like raspberry pi) which will also act as an interface for the Audio-visual
Design and Development of a Driving Assistance 37

alert and HUD system. The pothole detection is entirely based on Image
processing and computer rather than machine learning to reduce the latency
and make the system as real time as possible. The driver’s attention moni­
toring is done using a deep learning approach implemented using Tensor
flow (an open source machine learning library), which will monitor the
gesture of the driver and will decide whether he/she is paying attention to
driving or not. The algorithms and techniques are implemented using python
3.7 as the choice of programming language and Open CV (an open source
computer vision library) and Tensor flow libraries.
The outline of this chapter consists of six sections. Section 3.1 illustrates
the introduction and motivation of the proposed work. Section 3.2 of this
chapter discusses the glimpse of state of art exists in the proposed area.
Section 3.3 discusses the proposed system, and workflow of the model.
Section 3.4 emphasis on the results and related discussions. Finally, Section
3.5 concluded the proposed work.

3.2 EXISTING APPROACHES

To contextualize this work, several previous studies are outlined on the

detection of potholes and distress on the road. There are various approaches
to detect and identify the potholes on the road. Kang, Byeong Ho Choi, and
Su Il developed a pothole tracking system and method through 2D LiDAR
and Camera (Kang and Choi, 2017). They joined various heterogeneous
sensor systems to improve the accuracy of the overall system. Information
regarding angle and distance can be obtained by LiDAR (light detection and
ranging). The algorithm of pothole detection consists of removing noise,
pre-processing, clustering, line segment extraction and gradient of pothole
data function. This approach is very useful but not very cost effective due to
added cost of multiple heterogeneous sensors. Another way is to solely use
image processing to identify and locate the potholes in the image. Images
are captured by a camera mounted on the dashboard of the moving vehicle
and then pre-processed on the basis of difference of Gaussian-Filtering and
image segmentation methods based on clustering as proposed by Vigneshwar
and Kumar (2016). Based on various performances, it was observed that the
segmentation based on the K-Means clustering is desirable due to its fastest
computing time whereas edge segmentation based on edge detection is desir­
able for its specificity. Computer Vision based techniques such as seeing the
appearance and shape of the potholes uses histograms of oriented gradients
38 Advances in Data Science and Computing Technology

(HOG) features being of pothole or not. Normalized graph cut segmentation

scheme is used to locate the pothole in the detected pothole images (Azhar
et al., 2016). To monitor the driver’s attention some techniques, divide the
entire view which the driver sees while driving (Windshield, speedometer,
ORVMs, etc.), into various sectors (seven sectors mostly) and eye/gaze
tracking to see where the driving is paying the most attention (Mizuno et
al., 2017). Other ways include using video processing and computer vision
to retrieve the information from changing environments or the monitoring
of the physical state of human operators. It uses the head orientation of the
driver to get the state of visual attention of the driver and the smart phone’s
camera is used to collect the video data for this process (Mihai et al., 2015).
Choudhury et al. (2013, n.d.) exploits object recognition towards recognition
of number plate of the vehicle running over the road.

3.3 PROPOSED SYSTEM MODEL

The system is divided into two independent parts:

• pothole detection; and
• driver’s attention tracking.
The two independent systems use different input sources (Cameras)
but a common processing environment to process the inputs and make the
decisions. The quality and performance of the camera use will directly
affect the overall functioning of the system. A decent camera (preferably
night vision enabled) is required. All the input images are resized to 1280
× 720 pixels to make the processing faster without compromising on the
efficiency and accuracy of the system. The pothole detector and the driver’s
attention monitor will run in parallel using multiprocessing techniques on
the same onboard processing environment (a SOC) and will be controlled
by a decision making algorithm that will decide when to alert the driver. The
driver can be alerted using multiple Ways, but it is observed that Visual clues
are the easiest to detect by the human brain, for example flashing lights, or
blinking LEDs, etc.
This system model described in Figure 3.1 shows the two independent
systems working in parallel. The pre-processing for both the systems is quite
different as they work in different manner as shown in the next section of
this chapter. As the driver’s attention monitor is based on Deep learning, it
performs significantly better on GPUs rather than CPU.
Design and Development of a Driving Assistance 39

FIGURE 3.1 Model of proposed system.

3.3.1 DRIVER’S ATTENTION MONITOR

The model that this chapter uses for the detection of the attention of the driver
is based on the image classification implemented using tensor flow. In this, a
deep learning model is trained to classify images among various classes and
then recognize them based on the training received. The state farm dataset
(Abouelnaga, Eraqi, and Moustafa, 2017) which contains snapshots from a
video captured by a camera mounted in the car is used in this project. The
training set has ~22.4 K samples with equal distribution among the classes
and 79.7 K unlabeled test samples.
There are 10 classifications in the state farm dataset:
• Safe driving;
• Texting using right hand;
• Talking on the phone using right hand;
• Texting using left hand;
• Talking on the phone using left hand;
• Operating radio/multimedia system;
• Drinking;
• Reaching behind;
• Hair and makeup;
• Talking to passenger(s).
The driver’s attention monitor was trained using tensor flow transfer
learning (Chris, n.d.). Inception v3 (It is a widely-used image recognition
model) (Szegedy et al., 2016) which is trained on Image net dataset (Deng
40 Advances in Data Science and Computing Technology

et al., 2009) is used as a starting point and retrained using transfer learning
to recognize the above mentioned classes of images. A 2048-dimensional
vector for each image is passed to the Top layer as an input. It trains a soft
max layer on top of this layer. Suppose the soft max layer contains N labels
(10 in our case), then this corresponds to training N + 2048*N (or 1001*N)
model parameters corresponding to the learned weights and biases. The
Deep learning model that was obtained after the training process was used to
test on the 10% of the training images, that is, 10% of 20,000, which is 2,000
images. The tested model was then re-assessed over number of iterations to
increase the accuracy and performance of the model. The testing at the last
step was done on over 4,000 images.

3.3.2 POTHOLE DETECTOR

The flowchart described in Figure 3.2 shows the technique and algorithm used.
It uses a mix of computer vision and image processing techniques to find the
potholes/distress in the input feed. Firstly, the given input frames are converted
from BGR to Gray scale to improve the performance simply because BGR
images have three channels and gray scale images have one channel. The Gray
Scale image is passed through a 3×3 Gaussian filter to remove noise before
passing it to canny edge algorithm to detect the edges in the image which
is the base of our classification (Canny, 1986). The detected edges serves 2
purposes, one is to find the lanes in the road to extract the region of interest
(ROI) from the road which helps to remove unnecessary parts from the images
which greatly improves the performance and reduces latency. The lanes are
detected using Hough Lane detection, which converts the edges into lanes
which are filtered to get the ROI. This system uses a 10 second timer to reset
the position of lanes so that it can counter changing road conditions. The ROI
is superimposed on the edges to find the contours only on the surface of the
road. These detected contours are filtered based on their shape and size to
avoid false detections. The filtered contours are then projected back.

3.4 RESULTS AND DISCUSSION

The system has been developed using AMD A8-APU running at 2.2 Ghz
with 8 GB of RAM for testing both the systems. Testing on the video feeds
gave us a speed of 30 Frames per second making it more real-time. In our
experimentation, we have found that our method with computer vision was
Design and Development of a Driving Assistance 41

more real-time. Computer vision methods are easy to implement and manu­
facture as well provide with better real-time results. The number of frames
one system can process in a second with desired accuracy was chosen to
measure the performance of both the systems. The results of the two models
are further explained in subsections.

FIGURE 3.2 Pothole detector algorithm.

3.4.1 DRIVER’S ATTENTION MONITOR

The driver’s attention monitor was tested on us in a live video stream in an

automobile. It predicted correct results in each of the 10 classifications on
which it was based. The efficiency of the model was around 68%. Figure 3.3
shows a few samples of result obtained from the driver’s attention model.
Figure 3.3(a) correctly depicts the user talking on phone with his right hand.
Figure 3.3(b) depicts the user reaching behind. Figure 3.3(c) depicts the user
is drinking something. Figure 3.3(d) shows that the user is careful.

3.4.2 POTHOLE DETECTOR

The pothole detector, when worked on a sample video feed, works efficiently
Processing 30 Frames per second on an average. Some of the result samples
are given below Figure 3.4 describes the process of pothole detection and the
output. Figure 3.4(a) is the input frame of the road image. Figure 3.4(b) shows
the initial ROI, that is, just the road, obtained from the frame of Figure 3.4(a).
Figure 3.4(c) depicts the output of the canny edge detector that is used to detect
42 Advances in Data Science and Computing Technology

the lane in the road as shown in Figure 3.4(d). Figure 3.4(e) further reduces the
final ROI to the driving lane at the same time, detects the edges for the pothole
contours. Figure 3.4(f) shows the bounding box around the pothole contour
that are being detected. This image is also being projected to the driver.

(a) (b)

(c) (d)
FIGURE 3.3 Output of driver’s attention model: (a) depicts the user talking on phone using
his right hand; (b) depicts the user reaching behind; (c) depicts the user is drinking something;
and (d) shows that the user is careful.

3.5 CONCLUSIONS

The approach and procedure suggested in this chapter has not been tested
practically in real life scenario, hence this just acts as a prototype system.
Design and Development of a Driving Assistance 43

(a) (b)

(c) (d)

(e) (f)
FIGURE 3.4 Output of pothole detector at various stages: (a) is the input frame of the road
image; (b) shows the initial region of interest (ROI); (c) depicts the output of the canny edge
detector that is used to detect the lane in the road as shown in (d); (e) reduces the final region of
interest (ROI) to the driving lane at the same time, detects the edges for the pothole contours;
and (f) shows the bounding box around the pothole contour that will are being detected.
44 Advances in Data Science and Computing Technology

If implemented properly, this technology can improve the quality of driving

making it easier and save lives as it reduces accidents. At the same time,
the techniques suggested in this chapter also help government agencies
around the world to improve and monitor the condition of roads. Computer
vision and machine learning are the future and consistent improvements are
being made in these fields with each passing day. We should seek more and
more ways to use these technologies in our daily lives to make an impact on
the society. Techniques such as computer vision, image processing, deep/
machine learning are the basic fundamental blocks for the artificial intelli­
gence on which the future will be shaped as it helps the society in improving
the quality of tasks as well as makes the task easier. Artificial intelligence
based on the above-mentioned technologies is already being used in variety
of fields such as medical, mining, robotics, space studies, industries, etc.,
and its uses will increase as the years pass by to various other fields. Hence,
this chapter suggests that this model should be used in automobile industries
so that it can provide various aforementioned to the users.

KEYWORDS

• computer vision
• deep learning
• image classification
• image processing
• object detection
• pothole detection

REFERENCES

Abouelnaga, Y., Eraqi, H. M., & Moustafa, M. N., (2017). Real-Time Distracted Driver Posture
Classification. No. Nips, [Online]. Available: http://arxiv.org/abs/1706.09498 (accessed on
08 December 2021).
Azhar, K., Murtaza, F., Yousaf, M. H., & Habib, H. A., (2016). Computer vision based
detection and localization of potholes in asphalt pavement images. In: 2016 IEEE Canadian
Conference on Electrical and Computer Engineering (CCECE) (pp. 1–5). doi: 10.1109/
CCECE.2016.7726722.
Design and Development of a Driving Assistance 45

Canny, J., (1986). A computational approach to edge detection. IEEE Trans. Pattern Anal.
Mach. Intell., PAMI-8(6), 679–698. doi: 10.1109/TPAMI.1986.4767851.
Choudhury, A., & Negi, A. (2016). A new zone based algorithm for detection of license plate
from Indian vehicle In: 2016 Fourth International Conference on Parallel, Distributed and
Grid Computing (PDGC) (pp. 370–374). IEEE.
Chris, D. (2018). Tensor flow Image Classification. https://github.com/Microcontrollers And

More/TensorFlow_Tut_2_Classification_Walk-through (accessed on 08 December 2021).

Deng, J., Dong, W., Socher, R., Li, L. J., Li, K., & Fei-Fei, L., (2009). ImageNet: A large-scale

hierarchical image database. In: CVPR09.

Guardian, T. (2018). More Deadly Than Terrorism: Potholes Responsible for Killing 10 People
a Day in India. https://www.theguardian.com/world/2018/jul/25/more-deadly-than-terrorism­
potholes-responsible-for-killing-10-people-a-day-in-india (accessed on 08 December 2021).
India Today. (2018). Over 9300 Deaths, 25,000 Injured in 3 Years Due to Potholes. https://
www.indiatoday.in/india/story/over-9300-deaths-25000-injured-in-3-years-due-to­
potholes-1294147-2018-07-24 (accessed on 08 December 2021).
Kang, B. H., & Choi, S. I., (2017). Pothole detection system using 2D LiDAR and camera. In:
International Conference on Ubiquitous and Future Networks (pp. 744–746). doi: 10.1109/
ICUFN.2017.7993890.
Mihai, D., Dumitru, A., Postelnicu, C., & Mogan, G., (2015). Video-based evaluation of driver’s
visual attention using smartphones. In: 2015 6th International Conference on Information,
Intelligence, Systems and Applications (IISA) (pp. 1–5). doi: 10.1109/IISA.2015.7387983.
Mizuno, N., Yoshizawa, A., Hayashi, A., & Ishikawa, T., (2017). Detecting driver’s visual
attention area by using vehicle-mounted device. In: 2017 IEEE 16th International
Conference on Cognitive Informatics Cognitive Computing (ICCI*CC) (pp. 346–352). doi:
10.1109/ICCI- CC.2017.8109772.
Roy, S., Choudhury, A., & Mukherjee, J., (2013). An approach towards detection of Indian
number plate from vehicle. International Journal of Innovative Technology and Exploring
Engineering (IJITEE), 2(4), 241–244.
Sharma, S. K., & Sharma, R. C., (2019). Pothole detection and warning system for Indian roads.
In: Advances in Interdisciplinary Engineering, Lecture Notes in Mechanical Engineering (pp.
511–519).
Szegedy, C., Vanhoucke, V., Ioffe, S., Shlens, J., & Wojna, Z., (2016). Rethinking the inception
architecture for computer vision. Proc. IEEE Comput. Soc. Conf. Comput. Vis. Pattern
Recognit., 2818–2826. doi: 10.1109/CVPR.2016.308.
Vigneshwar, K., & Kumar, B. H., (2016). Detection and counting of pothole using image
processing techniques. In: 2016 IEEE International Conference on Computational Intelli­
gence and Computing Research (ICCIC) (pp. 1–4). doi: 10.1109/ICCIC.2016.7919622.
CHAPTER 4

PREDICTION OF THE POSITIVE CASES

OF COVID-19 FOR INDIVIDUAL
STATES IN INDIA AND ITS MEASURE:
A STATE-OF-THE-ART ANALYSIS
AMITAVA CHOUDHURY,1 KAUSHIK GHOSH,1 and
SUMAN GHOSAL2
1
School of Computer Science, University of Petroleum and Energy

Studies, Dehradun, Uttarakhand, India,

E-mail: [email protected] (A. Choudhury)

2
Department of Electrical and Electronics Engineering, Pailan College

of Management and Technology, Kolkata, West Bengal, India

ABSTRACT

India too is been affected by the global pandemic of COVID-19. It is there­

fore, necessary to predict the number of cases well in advance such that
the country remains prepared well in advance to face the situation. Many
predictions have come up for the entire country, however, for a country
like India, it is more suitable to consider the situation of individual states
separately. In this chapter, we have thus predicted the number of expected
cases of COVID-19 for India along with a deeper analysis of data for majorly
affected states. The results have shown that a state wise analysis is going to
give better results for prediction with a lower margin of error, as compared to
a countrywide analysis. Moreover, we have also identified three best states
that have performed better than others in terms of controlling the rate of
transmission among people.
48 Advances in Data Science and Computing Technology

4.1 INTRODUCTION

The pandemic caused by COVID-19 virus has both shook and united the
world like never before. Never in the recent past, has any epidemic concur­
rently affected governments across the globe with the same sense of urgency.
With the global casualty figures crossing over 200,000 (worldometers.info/
coronavirus, 2020)—different countries and agencies are revisiting their
warfare strategies against this virus. Originated in China (Huang et al.,
2020), the pandemic has caused havoc in other countries like Italy and USA
as well (Graselli et al., 2020). The entire Europe and North America for that
matter are yet to control the pandemic. India too is fighting her war against
this pandemic. Figures 4.1(a) and (b) have shown the number of active cases
of COVID-19, worldwide, and India respectively.

FIGURE 4.1 (a) Active cases worldwide (Worldometers, 2020); (b) active cases with in
India (Wikipedia, 2020).

The first case of Novel Corona virus (COVID-19) in India was reported
on 30th January, 2020 (PIB, 2020). WHO declared it as a global pandemic
(Wang et al., 2020) and as a result the Indian government declared an initial
lockdown of 21 days (Advisory Trade Facilitation, GoI, 2020) on 24th of
March, 2020, in order to prevent community transmission. On 14th April,
India entered into the second phase of lockdown (till 3rd May, 2020), adhering
to the advice of different agencies. Different medical and administrative
bodies in the country (and outside too) proposed social distancing as one of
the means to deal with the pandemic (Singh and Adhikari, 2020).
As the end of second phase of lockdown in India is approaching, the
country has an approximate 30,000 (mygov.in, 2020) reported cases of
COVID-19. Although this is a large number and the numbers are expected to
increase further, yet it must be said that in a country like India, with a popu­
lation more than 1.35 billion (Wikipedia, 2020) and a population density
Prediction of the Positive Cases of COVID-19 49

of 464/Km2 (worldometers.info, 2020), the country has not yet entered into
the stage of community transmissions. Even if India is lucky enough to
avoid community transmissions, the total infected population is sure to be a
mammoth one due to the reasonably high population density of the country.
As a result, a prediction mechanism with an acceptable degree of accuracy is
required to estimate beforehand the number of people who might be infected
by the virus. This is required in order to keep both the government and the
health workers prepared for the upcoming months, such that they do not get
overwhelmed with the number of cases all of a sudden.
Tomar and Gupta (2020) have done predictions of the number of affected
patients countrywide, for a period of 30 days. However, a monolithic predic­
tion of this type is not appropriate for a country like India where the number
of infected persons vary by a huge margin for different states. To mention,
Maharashtra has the highest number of affected persons and Sikkim has
none.
For example, till 26th April, 2020, the total number of reported cases
in Maharashtra was more than 8,000, whereas, for at least 12 states/union
territories the number of confirmed cases were yet to reach the three-figure
mark (covid19india.org, 2020). One of the reason for this is a huge variation
in population density among the states of the country. In our results, we have
shown why a countrywide prediction will give a greater margin of error as
compared to predictions done for the individual states.
In this chapter, we have predicted the number of confirmed/active cases, for
a period of next 30 days, for the entire country. For the purpose, we have used
algorithm comprising of long short-term memory (Hochreiter and Schmid­
huber, 1997) and have compared the results for the major states and union
territories of India. On analyzing the results, we have found that different
states are performing differently while combating the current situation.
Our main contributions in this chapter are as follows:
• Predicting the total number of COVID-19 cases in India for coming
30 days;
• Showing that a countrywide prediction is bound to give greater margin
of error when compared to state wise prediction; and
• Identifying the best states in terms of preventive actions taken against
the outbreak of the pandemic.
We have trained our system with data ranging from 30th January, 2020 to
20th April, 2020. Thereafter, we have validated our results for the period of
21st April, 2020 to 25th April, 2020.
50 Advances in Data Science and Computing Technology

The prediction made in this chapter is for the period of one month, i.e.,
26th April, 2020 to 26th May, 2020. The rest of the chapter is organized in the
following way: Section 4.2 contains methodology of our work. Section 4.3
contains the results along with the data sets and finally we have concluded
in Section 4.4.

4.2 METHODOLOGY

The exact reason for the occurrence of COVID-19 is yet to be known and
hence it is quite difficult to categories the COVID-19 symptoms. As per
the guidelines shared by World Health Organization (WHO), COVID-19
symptoms are very similar to normal flu or influenzas. Hence, to create a
mathematical model for predicting the number of affected persons by the
virus is really difficult. In scenarios like this, recurrent neural network
(RNN) algorithm is very useful. Generally, this algorithm is used to predict
time series analysis based on large historical information. However, the
main disadvantage of such a model is the time consumed in training a large
dataset. This is due to the gradient exploding problem (Jiang and Schotten,
2020; Connor et al., 1994). As a solution, long short term memory (LSTM)
is often very useful and this model is having the capabilities to learn order
dependence in sequential prediction problems (Tomar et al., 2020; Hochre­
iter and Schmidhuber, 1997).

4.2.1 LONG SHORT TERM MEMORY (LSTM)

As discussed, LSTM has a similar process flow mechanism as RNN.

However, the difference is in the operations within the LSTM internal struc­
ture. The key component of LSTM consists of cell state. It is the transport
highway and is used to transfer relative information to all other blocks.
The other components are various gates, such as forget gate, input gate and
output gate.
Gates contains sigmoid activations as shown in Figure 4.2. A sigmoid
activation is similar to the tanh activation. Instead of squishing values
between –1 and 1, it squishes values between 0 and 1. That is helpful to
update or forget data because any number multiplied by 0 is 0; causing
values to disappear or to be “forgotten.”
Prediction of the Positive Cases of COVID-19 51

FIGURE 4.2 Sigmoid squishes values to be between 0 and 1.

A sigmoid function can be calculated as:

1
ϕ (z) = (1)
1+ e − z
1. Forget Gate: The functionality of this gate to decide whether an
information is to be thrown away or kept. Information from the
previous hidden state and information from the current input is
passed through the sigmoid function. Predicted values come out
between 0 and 1. Closer the value to 0, it is to forget, and closer it is
to 1, it is to be kept. It is formulated in Eqn. (2):
=f t σ (W f *[ ht −1 , X t ] + b f ) (2)

2. Input Gate: To update the cell state, input gate is used. First, the
previous hidden state and current input are pass into a sigmoid func­
tion as. That decides which values will be updated by transforming
the values between 0 and 1 which indicates not important and impor­
tant measure respectively. A hidden state and current input also pass
into the tanh function to squish values between –1 and 1 to regulate
the network. It is formulated in Eqn. (3).
52 Advances in Data Science and Computing Technology

=it σ (Wi *[ ht −1 , X t ] + bi ) (3)

3. Cell State: Initially, the cell state is multiplied by the forget vector.
This has a possibility of dropping values in the cell state if it is multi­
plied by values nearer to 0. Then we take the output from the input
gate and do a pointwise addition, which updates the cell state to new
values that the neural network finds relevant.
= (
Ct σ f i * Ct −1 + it * Ct ) (4)

4. Output Gate: The next hidden stated would be identified in this

stage which contains the information on previous inputs. It is used
for predictions as well. First of all, the previous hidden state and the
current input is passed into a sigmoid function. Thereafter, the newly
modified cell state is passed to the tanh function. On multiplying the
tanh output with the sigmoid output, it is decided which information
the hidden state should carry. The final output obtained is the hidden
state. The new cell state and the new hidden state is to be carried over
to the next time step.
=Ot σ (Wo *[ ht −1 , X t ] + bo ) (5)

4.3 RESULTS AND DISCUSSION

In this chapter, we used COVID-19 pandemic dataset collected from official

website of Govt. of India (COVID-19.in, 2020). To implement the work, we
have used python programming with few useful libraries such as pandas for
data preprocessing, scikit learn machine-learning analysis, matplotlib, and
seaborn for plotting data. The data has been collected from 30th January, 2020
to 20th April, 2020 and validated for the period 21st April, 2020 to 25th April,
2020. In this study, we tried to predict the probability of getting infected by
COVID-19 virus. Along with this, we tried to get a better understanding of
the effect of lockdown and other precautionary measures on the number of
reported cases. Firstly, we made the predictions for entire India and then,
in order to get an in-depth understanding, we did a state wise prediction
and analysis. We have also discussed the effect of the lockdown and other
measures in 12 major states of India. Figure 4.3 shows the comparison
between predicted value and actual value of confirmed cased for the period
21st April 2020 to 25th April 2020, for entire India.
Prediction of the Positive Cases of COVID-19 53

FIGURE 4.3 COVID-19 confirm cases actual vs. predicted from 21st April 2020 to 25th
April 2020.

A similar study has also been conducted by Tomar et al. (2020).

However, India is a country with a population more than 1.35 billion, with
wide variation in population density among different states. Hence, it is very
difficult to equate the condition of different states and predict the number of
affected persons. Hence, a state wise prediction and analysis is also essential.
Figure 4.4 has shown the state wise prediction results along with validation.
From Figure 4.4, it is clearly visible that the prediction results are varied
from state to state. In Figure 4.4(a, h, l), the prediction results are very close
to the actual report given by the government. It can therefore, be inferred
that these three states (Andhra Pradesh, Gujarat, and West Bengal) have
taken adequate medical measures and people of these states have followed
the lockdown in a rigorous manner. This ensures that the state government
has been able to reduce the rate of transmission by taking proper measures.
Whereas in Figure 4.4(c, d, e, f) the actual number of cases have superseded
the predicted results. Hence, it may be concluded that these states, viz.
Haryana, Maharashtra, Tamil Nadu, and Uttar Pradesh still need to put more
efforts in controlling the pandemic. May be an extension of the lockdown
period for these four states is recommendable in braking the transmission
chain of the infection. A tabular validation of the prediction results is given
for entire India and West Bengal respectively in Tables 4.1 and 4.2, along
with the percentage of error.
54 Advances in Data Science and Computing Technology

(a) (g)

(b) (h)

(c) (i)

(d) (j)
FIGURE 4.4(a–l) (Continued)
Prediction of the Positive Cases of COVID-19 55

(e) (k)

(f) (l)
FIGURE 4.4 (a–l) State wise COVID-19 cases prediction and validation between April 20th
2020 to April 25th 2020. Particularly for Gujarat we can see that the number of actual cases is
less than that of the predicted cases.

TABLE 4.1 Comparison of Reported and Predicted Cases for India

Date Actual Data Cited in Predicted Data Error Percentage
Government Website
21st April 2020 18,601 17,377 6.58
22nd April 2020 19,984 18,713 6.36
23rd April 2020 21,393 20,096 6.06
24th April 2020 23,077 21,505 6.81
25th April 2020 24,506 23,589 5.37

TABLE 4.2 Comparison of Reported and Predicted Best Case (West Bengal)
Date Actual Data Cited in Predicted Data Error Percentage
Government Website
21st April 2020 392 391 0.255102
22nd April 2020 423 429 –1.41844
23rd April 2020 456 466 –2.19298
24th April 2020 514 501 2.529183
25th April 2020 571 535 6.304729
56 Advances in Data Science and Computing Technology

The prediction results are based on statistical analysis. From Table 4.1,
it is also observed that from 21st April 2020 onwards the predicted cases
follows a linear pattern, which indicate a stability in number of effected
in India. It is now very crucial to decide whether to extend the lockdown
beyond 3rd May, 2020. As per our results, it can be said that as certain states
are doing better than others in terms of controlling the rate of transmission,
a phased removal of lockdown is recommendable for the country. Hence, we
can say that, within a month India can control the effect of COVID-19 and
will definitely back the normal day-to-day work.

4.4 CONCLUSION

From the results, it is clear that predicting the number of transmissions for
the entire country could give greater error percentage in the predicted data.
Therefore, it is recommended to go state wise in order to predict the number
of cases for an upcoming period of 30 days.
Analyzing the results of the individual states, we conclude that Gujarat,
West Bengal, and Andhra Pradesh have been able to keep the rate of trans­
mission lower as compared to any other state.
On the other hand, in states like Haryana, Maharashtra, Tamil Nadu and
Uttar Pradesh the rate of transmission of the infection has increased than
expected. May be an extension of the lockdown period could reduce the rate
of transmission in these states.

KEYWORDS

• COVID-19
• digital India
• long short-time memory
• machine learning
• pandemic
• prediction
Prediction of the Positive Cases of COVID-19 57

REFERENCES

Advisory, Trade Facilitation, (1945). Government of India. 2017–2020.

Connor, J. T., Martin, R. D., & Atlas, L. E., (1994). Recurrent neural networks and robust time
series prediction. IEEE Trans. Neural Netw., 5(2), 240–254.
Grasselli, G., Pesenti, A., & Cecconi, M., (2020). Critical care utilization for the COVID-19
outbreak in Lombardy, Italy: Early experience and forecast during an emergency response.
JAMA.
Hochreiter, S., & Schmidhuber, J., (1997). Long short-term memory. Neural Comput., 9(8),
1735–1780.
Huang, C., Yeming, W., Xingwang, L., Lili, R., Jianping, Z., Yi, H., Li, Z., et al., (2020).
Clinical features of patients infected with 2019 novel coronavirus in Wuhan, China. The
Lancet, 395(10223), 497–506.
Jiang, W., & Schotten, H. D., (2020). Deep learning for fading channel prediction. IEEE Open
J. Commun. Soc., 1, 320–332.
MyGov. (India), (2020). https://www.mygov.in/covid-19/ (accessed on 08 December 2021).
PIB, (2020). https://pib.gov.in/pressreleaseiframepage.aspx?prid=1601095 (accessed on 08
December 2021).
Singh, A. K., Singh, A., Shaikh, A., Singh, R., & Misra, A., (2020). Chloroquine and
hydroxychloroquine in the treatment of COVID-19 with or without diabetes: A systematic
search and a narrative review with a special reference to India and other developing
countries. Diabetes & Metabolic Syndrome: Clinical Research & Reviews.
Singh, R., & Adhikari, R., (2020). Age-Structured Impact of Social Distancing on the
COVID-19 Epidemic in India. arXiv preprint arXiv:2003.12055.
Tomar, A., & Gupta, N., (2020). Prediction for the spread of COVID-19 in India and
effectiveness of preventive measures. Science of the Total Environment, 138762.
Wang, L. S., Wang, Y. R., Ye, D. W., & Liu, Q. Q., (2020). A review of the 2019 Novel
Coronavirus (COVID-19) based on current evidence. International Journal of Antimicrobial
Agents, 105948.
Wikipedia, (2020). https://en.wikipedia.org/wiki/ (accessed on 08 December 2021).
Wikipedia, (2020). https://en.wikipedia.org/wiki/Demographics_of_India (accessed on 08
December 2021).
Worldometers, (2020). https://www.worldometers.info/coronavirus/ (accessed on 08 December
2021).
Worldometers, (2020). https://www.worldometers.info/world-population/india-population/
(accessed on 08 December 2021).
CHAPTER 5

DEEP RESIDUAL NEURAL NETWORK

BASED MULTI-CLASS IMAGE
CLASSIFIER FOR TINY IMAGES
ALOK NEGI,1 PRACHI CHAUHAN,2 and AMITAVA CHOUDHURY3
1
National Institute of Technology, Uttarakhand, India
G.B. Pant University of Agriculture and Technology, Pantnagar,
2

Uttarakhand, India
3
University of Petroleum and Energy Studies, Dehradun, India,
E-mail: [email protected]

ABSTRACT

With the rapid growth of digital object recognition in over the year, automatic
image classification and detection has become the most difficult challenge in
computer vision areas. Contrary to human experiences, automated percep­
tion, and interpretation of tiny images by machine is difficult. Numerous
researches have been conducted throughout the existing identification
model to address problem, but still the performance has been restricted.
Nevertheless, those approaches lack appropriate classification method. In
this proposed work, deep learning algorithms is used to obtain the predicted
outcomes in the field, such as computer visions and features a deep learning
algorithm often used to classify the images automatically. In order to classify
tiny images, we use the image-net data set as a standard. The tiny images
need more computational power to classify images. By training the images
using deep network we get the better accuracy response with in experimental
phase it demonstrates that our approach provides the high precision in image
60 Advances in Data Science and Computing Technology

classification. Accuracy and loss curve based qualitative analysis performed

for proposed work.

5.1 INTRODUCTION

Image recognition and identification are most important issues in computer

vision. These are the origin of several other complex vision problems, along
with optimization, detection, and action interpretation. Image classification
and identification (Barbedo, 2018a, b) are widely used throughout the several
areas, including facial recognition and intelligent video processing, image
classification, traffic scenario for vehicles in traffic counting, regressive
motion sensors and so on. The purpose of this proposed research is to
create various neural networks that classify different tiny images (Bastidas
2017) in advanced convolution. In deep learning, we considered the neural
networks which recognize the image on the basis of its characteristics.
Deep learning depends on large quantities of high-quality analysis to model
future patterns and trends of behavior. The data sets must be descriptive of
the classes we expect to forecast otherwise the method will generalize the
distribution including its weighted classes as well as the bias will ruins the
model. The most common methods of deep learning models are Convolution
neural network (Schmidhuber, 2015), recurrent neural network (RNN), auto
encoders, etc.
Convolutional neural networks (CNNs) that are nowadays considered
narrowly the dominant method just for object classification. The deep CNN
are trained by tuning the parameters of the network (Alom et al., 2018) to
enhance linking during the training process and performs conscientiously
in tiny image classification. In computer vision concept, CNNs were
known to the strong visual models which produce feature hierarchies that
allow for accurate and remarkable segmentation and also execute predic­
tions comparatively faster than other more algorithms, and simultaneously
retaining competitive performance.
In bygone models of image classification, there is a use of actual raw
pixels to classify or identify the images. Images can be categorized by color
histogram and edge detection. This approach has been exclusively successful
but before more complex variants are encountered. That is where the classical
image recognition keeps failing, because some other features are not taken
into consideration by the model and that is why consider convolutional neural
network (ConvNet). CNN is a type of model of the neural network that helps to
extract feature or higher image representations (Srivastava et al., 2014). CNN
Deep Residual Neural Network Based Multi-Class Image 61

takes the raw pixel data from the image, trains the model, and then automatically
extracts features such as margins. Rashad et al. (2011) highlighted patterns
from the image for better classification.

5.2 LITERATURE REVIEW

Chandrarathne et al. (2020) conducted a systematic analysis evaluating the

efficiency of classifying CNN for limited data. This research is performed on
two datasets that are publicly accessible and tests the efficiency of two key
approaches: scratched training on transfer learning. Classification prediction
accuracy of scratch training are calculated by adjusting the number of
convolutional layers for various architectures. Image classification efficiency
is calculated based on the number of fine-tuned layers tested. Based on this
analysis, we can infer that the better way to identify the tiny data is by fine
tuning. It gives high precision because the training and test data are quite
close. Furthermore, researchers reported that by learning the target layers with
a far lower learning rate, accuracies of small datasets can indeed be increased.
For many real world applications, those findings would be very beneficial.
Liu et al. (2020) proposed a deep learning method dependent on stack-
sparse coding, that incorporates the concept of sparse expression into the
machine learning network layout and the extensive use of sparse interpretation
of better multidimensional linear data decomposed capability and deep
structural benefits of multilayer nonlinear transformation. This addresses the
problem of approximating complex operations and builds a deep learning
approach with the potential to adapt. Alternatively, a sparse representation
classifier is proposed to optimize kernel functions to address the issue of poor
classification performance in learning models. The whole sparse classifier
for representation will enhance the precision of the classification of images.
Xin et al. (2019) explained that deep ConvNet has been used to classify
images that are variable in scaling, localization, and other types of distortion.
The learning network can obtain parallel processing functionality by W, and
apparently its parameters and high computational are lower than that of
the conventional neural network. Based on the findings of the error back-
propagation algorithm, authors suggested an advanced deep neural network
(DNN) training standard for a minimal classification loss at maximum
intervals. The cross entropy and M3CE are simultaneously evaluated and
merged to achieve better results. The analytical results demonstrated that M3
CE can improve cross-entropy, which is an important addition to the criteria
of cross-entropy. M3CE-CEc performed well in both databases.
62 Advances in Data Science and Computing Technology

Antun et al. (2019) concluded that the new approach of learning recon­
struction algorithm for reconstruction of images in the inverse problem
produces unstable approaches usually through deep learning. In addition,
the test shows growing anomalies of uncertainty, obstacles, and novel ideas
towards science. Specifically: minor disruptions result in a variety of different
objects. Various networks generate different artifacts and instabilities. Range
of failure to restore massive changes so there is a broad range of instabilities
as regards structural changes.
Shiddieqy et al. (2017) implemented a deep convnet raspberry pi plat­
form by using TensorFlow. Larger network size can increase model accuracy
rate. This whole system is capable of classifying two cat and dog types
which have several similarities. Raspberry pi computing capacity capable
of deploying 5 layer CNN network but still not adequate for classification of
applications in real time.
Tripathi et al. (2019) presented functional execution of image identifi­
cation to use a small, CNN, offering less complexity and providing better
classification precision for all sets of data analyzed. Automatic data image
processing fills the distance between image files at the human vision and the
pixel unit. Deep ConvNet are being expressively deployed for the analysis,
detection, and classification of images for a variety of tasks. Such neural
networks, analogous to the human algorithm, contain neurons with learning
set of weights that are learned to recognize and categorize various objects or
functionalities throughout the image.

5.2.1 PROPOSED WORK

The aim of proposed work is to classify tiny images among 200 classes
with less computational power using deep residual network ResNet18 and
ResNet50. Pre-trained ResNet18 and ResNet50 model are used as a feature
extractor for transfer learning and their comparative analysis is done for
finding the best solution based on Accuracy curve, Loss curve and confusion
matrix.

5.2.2 DATASET DESCRIPTION

A dataset that includes images of around 100,000 belonging to 200 classes at

several significant growth stages has been used. Dataset consist training and
validation set with 90,000 and 10,000 images respectively.
Deep Residual Neural Network Based Multi-Class Image 63

5.2.3 DATA PREPROCESSING AND AUGMENTATION

Images are transformed from BGR to RGB because python expects (R, G,
B) using the pre-process input function by default, and images are resized
to prevent data leakage issues because most images look identical. Data
augmentation is a technique that allows the data diversity available for training
models to be substantially expanded, without actively collecting new data.

5.2.4 MODEL IMPLEMENTATION

ResNet is a strong backbone model which is very widely used in multiple

computer vision tasks and developed by research team from Microsoft.
Model design is focused on the principle of residual blocks that enable
use of shortcut connections. Throughout the network architecture, these
are basically connections where all the input is kept as it is (not weighted)
and conveyed to a deeper layer. ResNets (Prabhu, 2018) have varying sizes
depends on how broad any of the model layers are and how many layers it
has, such that ResNet-18, ResNet-34, ResNet-50, ResNet-101, ResNet-110,
ResNet-152, ResNet-164, ResNet-1202. A deep residual (ResNets) network
consists of several stacked “Residual Units.”
In a simplified form, each unit can be represented by Eqns. (1) and (2):
=yi h ( xi ) + F ( xiWi ) (1)
xi +1 = f yi (2)
where; xi and xi+1 are input and output of the ith unit, and F is a residual
function with h(xi) = xi is an identity mapping (Labach et al., 2019) and f is
a ReLU function.

5.2.5 TRAINING AND FINE-TUNING

Architecture to random weights is initialized and trained for a number of

epochs. The model recognizes attributes from data with each epoch. The
training was using the stochastic momentum gradient descent (SGDM)
optimizer, with an initial learning rate of 0.001, a mini batch size of 100
and 15 epochs overall. The performance of evaluated model is assess using
accuracy as shown:
Accuracy = (TP + TN) / (TP + TN + FP + FN) (3)
64 Advances in Data Science and Computing Technology

5.3 RESULTS AND EVALUATION

In ANACONDA 3.0 the method was put into practice. A standard PC fitted
with 2.30 GHz Intel(R) Core(TM) i5-3567 CPU, 64-bit operating system, 8
GB RAM and 2 GB AMD Radeon R5 M330 graphics engine run on Windows
10 operating system has been configured throughout the entire training and
testing phase of the model mentioned significantly. As the activation function,
SoftMax, and ReLU are used and Adam (Adaptive Moment Estimation) is the
optimizer. Accuracy curve, loss curve, and confusion matrix based analysis
are per formed on the training and validation set of deep ResNet model.
Logarithmic loss of multiple classes also widely recognized as the
categorical cross entropy is used as a metric for this work. A perfect classifier
gets the log loss of 0 and is calculated by Eqn. (4).
logloss = 1 / N ∑ i ∑ yij log ( pij )
N M
j
(4)
The first experiment on the ImageNet dataset performed with ResNet18
for 15 epochs. The training accuracy recorded 89% with logarithm loss 0.515
and the validation accuracy was 53% with logarithm loss 1.939. Figures 5.1
and 5.2 show the accuracy and loss curve for the experiment 1.

FIGURE 5.1 Training and validation accuracy of ResNet-18.

The second experiment performed with ResNet-50 for 15 epochs. The

training accuracy recorded 99% with logarithm loss 0.084 and the validation
Deep Residual Neural Network Based Multi-Class Image 65

accuracy was 63% with logarithm loss 1.645 (Figure 5.3). Figures 5.4 and
5.5 shows the accuracy and loss curve for the experiment 2.

FIGURE 5.2 Training and validation loss of ResNet-18.

FIGURE 5.3 Confusion matrix of ResNet-18.

As per experimental results, validation accuracy and loss for ResNet50

recorded better results than ResNet18 in just only 15 epochs. The most
difficult part of this work was to train such type of huge dataset on 2.30 GHz
Intel(R) Core(TM) i5-3567 CPU, 8 GB RAM and 2 GB AMD Radeon R5
M330 graphics engine local machine. Due to higher computational time, less
number of experiments was performed on this large dataset. However, we
recorded 63% accuracy with logarithm loss 1.645 just in 15 epochs which is
quite good (Figure 5.6).
66 Advances in Data Science and Computing Technology

FIGURE 5.4 Training and validation accuracy of ResNet-50.

FIGURE 5.5 Training and validation loss of ResNet-50.

5.4 CONCLUSION

Classification of tiny image always be a challenging task. It depends on

a detailed knowledge. In our system we developed comprehensive deep
learning models for the classification of images, predicated on advanced
CNN framework. A dataset has been used that includes images of around
Deep Residual Neural Network Based Multi-Class Image 67

100,280 distinct images actually belonging to 200 classes at several signifi­

cant stages. Our experimental findings showed how our deep-learning model
is capable of successfully categorizing various types of tiny images. While
proposed work does have better results, it is possible to increase the number
of epochs for better results. We hope that our proposed program will make a
significant contribution to the research in image processing.

FIGURE 5.6 Confusion matrix of ResNet-50.

KEYWORDS

• convolutional neural network

• ResNet18
• ResNet50
• tiny images

REFERENCES

Alom, M. Z., Taha, T. M., Yakopcic, C., Westberg, S., Sidike, P., Nasrin, M. S., & Asari, V.
K., (2018). The History Began from Alexnet: A Comprehensive Survey on Deep Learning
Approaches. arXiv preprint arXiv:1803.01164.
Antun, V., Renna, F., Poon, C., Adcock, B., & Hansen, A. C., (2019). On Instabilities of Deep
Learning in Image Reconstruction-Does AI Come at a Cost? arXiv preprint arXiv:1902.05300.
Barbedo, J. G. A., (2018b). Impact of dataset size and variety on the effectiveness of deep
learning and transfer learning for plant disease classification. Computers and Electronics
in Agriculture, 153, 46–53.
Barbedo, J. G., (2018a). Factors influencing the use of deep learning for plant disease
recognition. Biosystems Engineering, 172, 84–91.
Bastidas, A., (2017). Tiny ImageNet Image Classification. https://pdfs.semanticscholar.org/1b
0c/2ba54f7e2f3f5b3a2098721d36e6079d0382.pdf (accessed on 08 December 2021).
68 Advances in Data Science and Computing Technology

Chandrarathne, G., Thanikasalam, K., & Pinidiyaarachchi, A., (2020). A comprehensive study
on deep image classification with small datasets. In: Advances in Electronics Engineering
(pp. 93–106). Springer, Singapore.
Labach, A., Salehinejad, H., & Valaee, S., (2019). Survey of Dropout Methods for Deep Neural
Networks. arXiv preprint arXiv:1904.13310.
Liu, J. E., & An, F. P., (2020). Image classification algorithm based on deep learning-kernel
function. Scientific Programming, 2020.
Prabhu, (2018). CNN Architectures – LeNet, AlexNet, VGG, GoogLeNet and ResNet. https://
medium.com/@RaghavPrabhu/cnn-architectures-lenet-alexnet-vgg-googlenet-and-resnet­
7c81c017b848 (accessed on 08 December 2021).
Rashad, M. Z., El-Desouky, B. S., & Khawasik, M. S., (2011). Plants images classification
based on textural features using combined classifier. International Journal of Computer
Science and Information Technology, 3(4), 93–100.
ResNet, AlexNet, VGGNet, Inception, (2020). Understanding Various Architectures of
Convolutional Networks. Retrieved from https://cv-tricks.com/cnn/understand-resnet-alexnet­
vgg-inception/ (accessed on 08 December 2021).
Schmidhuber, J., (2015). Deep learning in neural networks: An overview. Neural Networks,
61, 85–117.
Shiddieqy, H. A., Hariadi, F. I., & Adiono, T., (2017). Implementation of deep-learning
based image classification on single board computer. In: 2017 International Symposium on
Electronics and Smart Devices (ISESD) (pp. 133–137). IEEE.
Srivastava, N., Hinton, G., Krizhevsky, A., Sutskever, I., & Salakhutdinov, R., (2014).
Dropout: A simple way to prevent neural networks from overfitting. The Journal of Machine
Learning Research, 15(1), 1929–1958.
Tripathi, S., & Kumar, R., (2019). Image classification using small convolutional neural
network. In: 2019 9th International Conference on Cloud Computing, Data Science &
Engineering (Confluence) (pp. 483–487). IEEE.
Xin, M., & Wang, Y., (2019). Research on image classification model based on deep convolution
neural network. EURASIP Journal on Image and Video Processing, 2019(1), 40.
CHAPTER 6

IMPROVED TDOA-BASED NODE

LOCALIZATION IN WSN: AN ANN
APPROACH
ARPAN CHATTERJEE, ASHISH AGAROWALA, and
PARAG KUMAR GUHA THAKURTA
CSE Department, NIT Durgapur, West Bengal, India

ABSTRACT

A new node localization method depending on time difference of arrival

(TDOA) concept is proposed in this chapter to obtain minimized error in
the prediction of the sensor node position in a network. A multi-layer, feed-
forward, back propagation neural network consisting of one input layer,
three hidden layers with a (64-32-16) node structure and an output layer is
used to develop the proposed model. Different cases have been introduced
to train the model by varying the number of anchor nodes and their location
along with the number of sensor nodes. Various results for each of the cases
mentioned earlier show the effectiveness of the proposed approach in terms
of the performance metrics namely root mean squared error (RMSE), average
accuracy and the average range error. The proposed approach obtains an
accuracy of 99.142% which outperforms the result of existing work.

6.1 INTRODUCTION

A wireless sensor network (WSN) (Patel and Kumar, 2018) is comprised of

sensor devices that can communicate information, gathered from a monitored
70 Advances in Data Science and Computing Technology

field, through wireless links. The sensor is used to collect the data regarding
both physical and environmental parameters, such as pressure, heat, light,
etc. The output of these sensors is transmitted to the base station (BS) for
further processing (Guha and Roy, 2018). In recent years, the efficient design
of a WSN has become a leading area of research due to its significant nature
of adaptability in diverse applications. One of the challenging applications in
this field is that of node localization (Zhang et al., 2017). It is a cooperative
process where nodes with known locations are used to localize nodes at
unknown locations. These nodes with known locations as well as unknown
locations are called anchor nodes and sensor nodes respectively.
As WSNs are penetrating into the industrial domain, many research
opportunities are emerging in the area of node localization. For example,
over the past decades, the safety measures adopted by the coal industry
(Shrawankar and Mangulkar, 2018) has been an important factor in
constraining its development around the world. It is known that coal mining
accidents result in loss of life and property. In this issue, mines environment
monitoring through node localization can reduce such losses and ensure
miners safety as far as possible (Srikanth, Kumar, and Rao, 2018).
In this context, a time difference of arrival (TDOA) based node localization
method is proposed in this chapter to obtain minimized error in the prediction
of the sensor node position. The proposed approach relies on the distance
between the anchor nodes and the sensor nodes. In order to obtain such
distance, the time difference between the data transmission by sensor node
and data reception by anchor node is considered. In this work, an artificial
feed-forward neural network with three hidden layers is used. To train the
model have been introduced by different cases by varying the number of
anchor nodes and their location along with the number of sensor nodes.
The rest of this chapter is organized as follows: Section 6.2 introduces a
literature survey for completeness of the work. Next, Section 6.3 presents the
system model. The proposed approach is discussed in Section 6.4. The results
are shown in Section 6.5. Finally, we conclude our work in Section 6.6.

6.2 LITERATURE SURVEY

A number of techniques have been proposed for node localization, which

can be divided into two broad, categories such as range based and range free
(Paul and Sato, 2017). The range based technique relies on distance measure­
ment between the nodes. Some of those in this domain are mainly using
received signal strength (RSS), TDOA (Tambe and Krishna, 2018). On the
Improved TDOA-Based Node Localization in WSN 71

other hand, range free techniques are dependent on hop based information.
In 2017, Shahra et al. in 2017 compared range free and range based localiza­
tion algorithms. The result of this work shows that range based algorithms
are more accurate than range free algorithms. In 2019 (Wu et al., 2019),
the authors stated that TOA and TDOA, among the range based positioning
methods achieve higher position accuracy to perform the measurement on
minimized load requirement for the single sensor node. Samanta et al. in
2018 discussed a node localization using ANN to manipulate the location
of the sensor nodes which obtains an accuracy of 98%. In order to improve
the accuracy further by minimizing RMSE and Average Range Error, a new
procedure for node localization is introduced here as discussed next.

6.3 SYSTEM MODEL

6.3.1 NETWORK MODEL

A WSN having N number of nodes in a square field of size (M × M) can

be viewed as a two dimensional (2D) coordinate system. Each node is
represented in a 2-axes layout as (x, y) where 0 <= x, y <= 50. These nodes
are homogeneous in nature. Let us assume that the number of anchor nodes
and sensor nodes are m and (N − m), respectively. The positions of the
anchor nodes are assumed to be fixed while the sensor nodes are randomly
deployed in the network. Let, the set of anchor node positions and the set of
sensor node positions are represented by A and S, respectively. These can be
represented below:
A = [(xi, yi) |0 ≤ xi, yi ≤ 50 & i ∈ [1, m]] (1)
S = [(xj, yj) | 0 ≤ xj, yj ≤ 50 & j ∈ [1, N − m]] (2)
In the proposed work, the values of m are considered as 3, 4, and 5 for the
sake of simplicity. The network with m = 3 and the set A = [(0,0), (25,50),
(50,0)] is shown in Figure 6.1. The values of A = [(0,0), (50,0), (0,50), (50,50)]
and A = [(0,0), (50,0), (0,50), (50,50), (25,25)] are considered for m = 4 and
m = 5, respectively.

6.3.2 USEFUL DEFINITIONS

In order to develop the proposed model and measure its efficiency, the
following definitions are introduced next:
72 Advances in Data Science and Computing Technology

FIGURE 6.1 Network representation with set A.

 Definition 1 (TDOA): It is defined as the time difference between

the data transmission by sensor node and data reception by anchor
node. It is denoted by the following expression:
TDOA = d/s (3)
where; ‘d’ is the distance between anchor and sensor node and ‘s’ is
the speed of data transmission by the sensor.
 Definition 2 (Residual): A residual is a measure of the prediction
in error which is nothing but the difference between the actual and
predicted distance of the sensor node from an anchor node. It is
denoted by the following expression:
e = (dpred− dactual) (4)
where; e is the residual, dpred and dactual are the predicted and actual
distance of sensor node from anchor node respectively.
 Definition 3 (RMSE): It is defined as the standard deviation (SD) of
the residuals. It highlights how spread out these residuals are. In the
context of the proposed work, such RMSE values can be obtained by
the following expression:
2
∑ ∑
N −m m
RMSE = =i 1 =j 1 ij
e (5)
Improved TDOA-Based Node Localization in WSN 73

where; eij is the residual of itℎ sensor node from the jtℎ anchor node.
 Definition 4 (Average Accuracy): It defines how correctly the
position of the sensor node is predicted. If the position is predicted
more accurately, the error in the distance would be less and subse­
quently higher average accuracy can be obtained. It is calculated in
percentage by using the following expression.
 e 
Accuracy  ∑
Average =  ×100 (6)
 d actual 
 
 Definition 5 (Average Range Error): Let us assume, there lies a
group of sensor nodes within a specific range from an anchor node.
Then the average range error for this specific range is defined as the
average of the errors in terms of distance for all of the sensor nodes
with respect to the anchor nodes. This can be obtained as follows:
Average Range Error = ∑̅|̅ ̅ di ̅ ̅ −̅̅ dpred
̅ |̅̅ (7)
where; di∈R and R can be [(0,10), (10,20),(20,30), (30,40), (40,50),
(50,60), (60,70)] in our proposed work.

6.3.3 MORPHOLOGICAL OPERATION BASED ENHANCEMENT

Morphological operations make use of structuring elements while enhancing

retinal artifacts. Disk-shaped masks are generally used among researchers in
biomedical image data enhancement. Tophat and bottom-hat transformations
are represented (Digital Image Processing, Gonzalez) in Eqns. (8) and (9),
respectively.
Tophat (I ) = (I – (I(open)b)) (8)
Bottomhat (I ) = ((I(close)b) – I ) (9)
where; b is structuring element and morphological operation (I(open)b) is
erosion of I by b, followed by dilation with b and I(close)b is dilation of I by
b, followed by erosion with b. According to Hassanpour et al. (2015), subse­
quent applications of tophat and bottomhat transformation lead to contrast
enhancement between light and dark arenas.
En = I + Tophat (I ) – Bottomhat (I ) (10)
When Eqn. (10) is tried with different structural elements of suitable size
enhancement result varies. The output enhanced retinal image is (En).
74 Advances in Data Science and Computing Technology

6.4 PROPOSED MODEL

In order to obtain an improvement on average accuracy and reduce RMSE

and average range error, the proposed work is carried out in two consecutive
steps. Initially, an ANN model has been developed in this regard. Next, the
developed model is trained with the input matrix consisting of the actual
distance between anchor and sensor nodes and the coordinate (x, y) of the
sensor nodes. For the purpose of testing the model, a new matrix consisting
of distance is fed to the model and subsequently the model predicts the coor­
dinate (X, Y) of the unknown sensor node.

6.4.1 ANN MODEL DEVELOPMENT

In the proposed work, a multi-layer, feed-forward, back propagation neural

network is used as shown in Figure 6.2. It consists of one input layer, three
hidden layers with a (64-32-16) node structure and an output layer. The
output layer consists of two nodes for the coordinate X and Y. Rectified
linear unit (ReLU) (Nwankpa et al., 2018) is used as the activation function
to cut off all the negative values of the node output. Here, the Adam optimizer
algorithm (Kingma and Ba, 2014) is used to adjust weight factors, which in
turn updates the output of the nodes in order to reduce error.

FIGURE 6.2 Feed forward neural network with one input, three hidden and two output layers.
Improved TDOA-Based Node Localization in WSN 75

6.4.2 MODEL TRAINING AND TESTING

There are six cases to be considered for training the model developed in the
earlier step. In each of the cases, the number of anchor nodes and sensor
nodes have been varied as listed in Table 6.1. The model has been trained and
tested separately for each of the six different cases.

TABLE 6.1 Various Cases with the Value of m and (N-m)

Case m N-m
1 3 100
2 3 500
3 4 100
4 4 500
5 5 100
6 5 500

The generation of a training dataset is required for training the model. It

is obtained by deploying the sensor nodes in the network area. For each of
the sensor nodes, the distance from all the anchor nodes is calculated. So, a
distance matrix D of size (N-m)*m is obtained. Each row in D represents a
feature vector as Vj = [Dj1, Dj2, …, Djm], where Dji denotes the distance of the
jth sensor node from the ith anchor node and Vj is the feature vector for the
jth sensor node. The original position (x, y) of the sensor node j is the label
of Vj. Each row of the input matrix consists of Vj along with its label for a
particular sensor node. The input matrix is then fed to the model which is
mentioned in Figure 6.3.

FIGURE 6.3 Input matrix used.

To test the model, a random set of 30 sensor nodes is generated in the

network area for each of those six cases introduced earlier. Then the feature
vector for each of these test sensor nodes is computed. Similar to the input
matrix used for training the model, each row of the test input matrix consists
 76 Advances in Data Science and Computing Technology

of the feature vector for a particular sensor node. This test input matrix is
then fed to the trained model which in turn predicts the label (X, Y) for each
of the test sensor nodes. Hence, the output of this model can be obtained as
follows:
X 
= Y  relu (relu (relu (D ∗ W1 + B1 ) ∗W2 + B2 ) ∗W3 + B3 ) ∗W4 + B4 (11)
 
where; D is the input row vector of length m; Wi is the weight vector at ith
layer and Bi is the bias vector at the ith layer.

6.5 RESULTS

In this section, a comparison among the six cases is discussed in an individual

and collective manner. A number of tests have been conducted using various
activation functions like sigmoid, tanh, ReLU, and by changing the number
of nodes in the different hidden layers. A network with three layers having
(64-32-16) node structure with ReLU as the activation function provides the
best result for the six cases.
It is noteworthy that 30 test sensor nodes have been generated in the
network area to check the prediction accuracy of the model. The predicted
position of these test sensor nodes is obtained after feeding the test input
matrix to the model. The results of the original and predicted position of test
sensor nodes for each of the different cases mentioned earlier are generated
as shown in Figure 6.4.
To compute the RMSE value, an error matrix of size (30 * m) is gener­
ated from the predicted position of the test sensor node. Here, an output
matrix of size (30 * m) is obtained. Each row of the output matrix is a
vector of size m denoting the distance of the particular sensor node from
the ‘m’ anchor nodes. The difference between the test input matrix and the
corresponding output matrix provides the error matrix. The corresponding
RMSE value is then computed as per case 5. On the other hand, the accu­
racy of a single test sensor node from a particular anchor node can be
obtained by considering the ratio between error in prediction and original
distance between the test sensor node and anchor node. There are (30 * m)
values obtained for all the test sensor nodes. The average of these values
denotes the average accuracy of the model as per case 6. So, a comparative
study on the RMSE value and the average accuracy for each of these cases
has been shown in Table 6.2.
Improved TDOA-Based Node Localization in WSN 77

(a) (b)

(c) (d)

(e) (f)
FIGURE 6.4 Original and predicted position of test sensor nodes for: (a) case 1; (b) case 2;
(c) case 3; (d) case 4; (e) case 5; (f) case 6.

The range error of a test sensor node provides an estimate of the error in
prediction of the position of the node when its distance from the anchor node
lies within a known range. The average range error of a particular range is
obtained as per case 7. In this context, from Figure 6.5, the range in which
the least average error is obtained. This information can be used to minimize
the error in prediction during the node localization process.
78 Advances in Data Science and Computing Technology

TABLE 6.2 RMSE and Average Accuracy (in %) for Different Cases
Cases RMSE Average Accuracy (in %)
1 0.973 97.004
2 0.564 98.193
3 0.759 98.015
4 0.287 99.142
5 0.57 98.476
6 0.362 98.902

 Remark 9.1: It is observed that our proposed work obtains an

accuracy of 99.142% over an accuracy of 98% which was obtained
by the existing work (Samanta et al., 2018). So, an improvement in
accuracy is obtained by our proposed work over existing work.

FIGURE 6.5 Average range error for different number of anchor nodes and sensor nodes.

6.6 CONCLUSIONS

A new approach of node localization in WSN is proposed in this chapter

to know the position of the sensor nodes for providing better communica­
tion among the nodes. Here, an ANN model with node structure (64-32-16)
is developed. The TDOA concept is used to develop the proposed model.
Different cases have been considered for training and testing the model
by varying the number of anchor nodes and sensor nodes in the network
area. Among these, the case with 4 anchor and 500 sensor nodes provides
Improved TDOA-Based Node Localization in WSN 79

maximum average accuracy (99.142%) and minimum RMSE (0.287). For

this case, the average range error as obtained is minimum in the range of
(20–30). This proposed model can be enhanced for heterogeneous nodes in
the near future.

KEYWORDS

• artificial neural networks (ANN)

• average accuracy
• average range error
• node localization
• root of mean square error (RMSE)
• time difference of arrival (TDOA)
• wireless sensor network (WSN)

REFERENCES

Guha, T. P. K., & Roy, S., (2018). A decentralized fuzzy c-means minimal clustering protocol
for energy efficient wireless sensor network. In: 2018 Fifth International Conference on
Parallel, Distributed and Grid Computing (PDGC) (pp. 24–29). Solan Himachal Pradesh,
India.
Kingma, D. P., & Ba, J., (2014). Adam: A Method for Stochastic Optimization. https://arxiv.
org/abs/1412.6980 (accessed on 08 December 2021).
Nwankpa, C., Ijomah, D., Gachagan, A., & Marshall, S., (2018). Activation Functions:
Comparison of Trends in Practice and Research for Deep Learning. https://arxiv.org/
abs/1811.03378 (accessed on 08 December 2021).
Patel, N. R., & Kumar, S., (2018). Wireless sensor networks’ challenges and future prospects.
In: 2018 International Conference on System Modeling & Advancement in Research Trends
(SMART) (pp. 60–65). Moradabad, India.
Paul, A. K., & Sato, T., (2017). Localization in wireless sensor networks: A survey on
algorithms, measurement techniques, applications and challenges. Journal of Sensor and
Actuator Networks, 6(4).
Samanta, R., Kumari, C., Deb, N., Bose, S., Cortesi, A., & Chaki, N., (2018). Node localization
for indoor tracking using artificial neural network. Third International Conference on Fog
and Mobile Edge Computing (FMEC) (pp. 229–233). Barcelona.
Shahra, E. Q., Sheltami, T. R., & Shakshuki, E., (2017). A comparative study of range-free and
range-based localization protocols for wireless sensor network: Using COOJA simulator.
International Journal of Distributed Systems and Technologies, 8(1).
80 Advances in Data Science and Computing Technology

Shrawankar, U., & Mangulkar, P., (2018). Monitoring and safety system for underground
coal mines. IEEE International Conference on Power Energy, Environment and Intelligent
Control (PEEIC2018).
Srikanth, B., Kumar, H., & Rao, K. U. M., (2018). A robust approach for WSN localization
for underground coal mine monitoring using improved RSSI technique. Mathematical
Modeling of Engineering Problems, 5, 225–231.
Tambe, K., & Krishna, M. G., (2018). An efficient localization scheme for mobile WSN.
International Journal of Innovative Technology and Exploring Engineering (IJITEE), 8(2S).
Wu, P., Su, S., Zuo, Z., Guo, X., Sun, B., & Wen, X., (2019). Time difference of arrival (TDoA)
localization combining weighted least squares and firefly algorithm. Sensors, 19, 2554.
Zhang, X., Tepedelenlioglu, C., Banavar, M., & Spanias, A., (2017). Node localization in
wireless sensor networks. In: Node Localization in Wireless Sensor Networks (Vol. 12,
pp. 1–62). (Synthesis Lectures on Communications). Morgan and Claypool Publishers.
 PART II

Role and Impact of Big Data in

E-Commerce and the Retail Sector

CHAPTER 7

IMPACT OF BIG DATA ON

E-COMMERCE WEBSITES

DIVYANG BHARTIA1 and MAUSUMI DAS NATH2

1
Student, St. Xavier’s College (Autonomous), Kolkata,
30 Mother Teresa Sarani, Kolkata–700016, West Bengal, India
2
Assistant Professor, St. Xavier’s College (Autonomous), Kolkata,
30 Mother Teresa Sarani, Kolkata–700016, West Bengal, India,
E-mail: [email protected]

ABSTRACT

There has been an increasing prominence in the analytics of E-Commerce

business lately. Nonetheless, it remains ineffectively explored as a theory,
which blocks its theoretical and useful advancement. This time dissimilar to
any is confronted with touchy development in the size of information created/
caught. Information development has gone through a renaissance, impacted
basically by ever less expensive figuring power and the omnipresence of the
web. The web has changed E-Commerce and client presently approach the
wide scope of items offered through E-Commerce sites. To stay serious and
protect a piece of the overall industry, E-Commerce firms figure web based
showcasing systems dependent on continuous information. This has directed
to a change in perspective in E-Commerce, where information is viewed
as the greatest resource for the firm in understanding explicit necessities
of clients, foreseeing conduct, fitting explicit requirements, and offering
execution measurements to survey viability. Internet business firms are
discovering approaches to remove important data from bigger datasets where
information gets created at more noteworthy speed, distinctive assortment,
84 Advances in Data Science and Computing Technology

and at high volumes that are regularly alluded to Big Data. Thus, this chapter
explains how E-Commerce firms are contributing gigantic to Big Data
Analytics to engage them to take precise and ideal choices.

7.1 INTRODUCTION

In the world of Digital and Electronic Commerce, advertisers take each

and every minute to catch the attention of the guests to fulfill their quick
needs. The key segment to fulfill client needs gives them the best client
experience which yields to two central points in business-brand value and
client dedication. To meet with the consumer loyalty, companies develop
huge information and research arrangements prompting accurately gauge
and break down each and every client execution variable on the site. Big
data refers to every single large volume of data that is been collected on
all aspects of human life. With Big Data, every action of humans could be
quantified and hence stored in the data bank which in turn assists in satisfying
the needs of the customers. These data are collected from a vast array of
electronic devices such as handheld devices like mobile, tablet, and even
portable devices such as laptops and all kind of sensors that are connected to
the internet. With Big Data, each activity of people could be evaluated and
thus put away in the information bank which thus helps with fulfilling the
necessities of the clients. This information is gathered from a huge swath
of electronic gadgets, for example, handheld gadgets like tablets and even
compact gadgets, for example, PCs, and all sort of sensors that are associated
with the web. Prior to Big Data companies applied investigation to reveal
patterns and increase bits of knowledge. In any case, the handling was
unmistakably complex and difficult to do physically. The personalization
should be possible by investigating the purchasing behavior of the client
that permits advertisers to set unmistakable patterns and examples. Extra
inputs permit the advertisers to adjust the endeavors and work on improving
contributions for the clients. Then again advertisers may think about
perusing information, buy styles and other large information sources to make
customized items. Google Trends online tools help advertisers to channel
down inclining points and related pursuit works permitting them to target
clients locally and all around including area, district, nations, and interests
to get more fascination towards the customized item or service These days
utilizing Big Data, advertisers likewise set ICP (ideal customer profile) that
permits them not to rely upon determined speculations but instead gives exact
Impact of Big Data on E-Commerce Websites 85

data about the socio-economics of the client just as their psychographic and
behavioristic examples. Consequently, Big Data permits advertisers to arrive
at clients at the ideal time and promote the item in a convincing manner to
understand their need and increment the deal.
The chapter is coordinated as follows: Section 7.2 details overview of
related work, Section 7.3 gives the exploratory study and Section 7.4 focuses
on the objectives of the study. Section 7.5 gives analysis and discussions
while Section 7.6 puts forward the recommendations and Section 7.7 finally
includes the references.

7.2 OVERVIEW OF RELATED WORK

The role of Big Data in marketing is increasing exponentially; there is no

comparison to the amount of data that is collected for the profiling of each
customer. To understand the aspects of the field few empirical studies have
been conducted to understand the role of big data analytics from the manage­
ment point of view. The use of data is a decisive variable in the marketing
decisions of the marketing managers and various related studies have been
conducted to understand the various fields of big data analytics from the
marketing perspective.
Shahriar and Samuel (2016) focused on more extensive conversations
with respect to future examination difficulties and openings in principle and
practice. Generally, the discoveries of the examination combine different
BDA ideas (e.g., the meaning of large information, types, nature, business
esteem, and important hypotheses) that give further experiences along with
the cross-cutting investigation applications in web-based business.
Uyoyo (2014) showed us an outline of the special highlights that
separate enormous information from customary datasets. Furthermore,
the use of large information investigation in the E-trade and the different
advances that make examination of shopper information conceivable is
talked about.
Avinash and Akarsha (2017) researches how the utilization of huge infor­
mation investigation is seen as worth maker that can direct E-Commerce
organizations accomplish upper hand.
Moorthi, Srihari, and Karthik (2017) talks about different approaches and
cycles continued in online business for business knowledge. Additionally,
proposed some new strategies to improve the business knowledge in web
based business field utilizing large information examination.
86 Advances in Data Science and Computing Technology

7.3 EXPLORATIVE STUDY

Walmart, with more than 20,000 stores in 28 nations is the biggest retailer on
the planet. So, it is fitting then that the organization is building the world’s
biggest private cloud, sufficiently large to adapt to 2.5 petabytes of informa­
tion consistently. To understand this data, Walmart has made what it calls
its Data Café – a best in class investigation center point situated inside its
Bentonville, Arkansas central station. The Data Café permits tremendous
volumes of interior and outer information, including 40 petabytes of ongoing
value-based information, to be quickly displayed, controlled, and pictured.
Walmart has an expansive huge information biological system. The large
information environment at Walmart forms numerous Terabytes of new infor­
mation and petabytes of authentic information consistently. The examination
covers a large number of items and 100’s of millions clients from various
sources. Walmart has comprehensive client information of near 145 million
Americans of which 60% of the information is of U.S. grown-ups. Walmart
accumulates data on what client’s purchase, where they live and what are the
items they like through in-store Wi-Fi. The enormous information group at
Walmart Labs examinations each interactive activity on Walmart.com-what
purchasers purchase available and, on the web, what is drifting on Twitter,
nearby occasions, for example, San Francisco goliaths winning the World
Series, how neighborhood climate deviations influence the purchasing
behaviors, and so forth. All the occasions are caught and broke down wisely
by huge information calculations to recognize important large information
bits of knowledge for a huge number of clients to appreciate a customized
shopping experience (Figure 7.1).
How Walmart is having a genuine effect to expand deals?
1. Launching New Products: Walmart is utilizing web based life
information to discover about the drifting items with the goal that
they can be acquainted with the Walmart stores over the world.
For example, Walmart dissected internet based life information to
discover the clients were distracted about “Cake Pops.” Walmart
reacted to this information investigation rapidly and Cake Pops hit
the Walmart stores.
2. Better Predictive Analytics: Walmart has as of late changed its
transportation approach for items dependent on large information
examination. Walmart utilized prescient examination and expanded
the base sum for an online request to be qualified with the expecta­
tion of complimentary delivery. As indicated by the new dispatching
Impact of Big Data on E-Commerce Websites 87

strategy at Walmart, the base sum with the expectation of compli­

mentary transportation is expanded from $45 to $50 with expansion
of a few new items to improve the client shopping experience.
3. Customized Recommendations: Simply the way wherein Google
tracks customized ads, Walmart’s enormous information calculations
break down MasterCard buys to give specific suggestion to its clients
dependent on their buy history.

FIGURE 7.1 An overview of how Walmart Labs function.

Source: slideshare.com.

7.3.1 WAL-MART VERSUS THE OPPOSITION

Despite the fact that business development at Wal-Mart has eased back as of
late, the measurements are still head and shoulders above the greater part of
the business. Contrasted with the best 10 retailers internationally, just Home
Depot beat Wal-Mart consequently on resources or net overall revenue. In
2013, Wal-Mart recorded a net edge of 3.5% and profit for resources of
8.6%. Costco, maybe its nearest rival, had an overall revenue of just 2% and
profit for resources of 6.8% that year. On a business premise, Costco utilizes
its land all the more proficiently, creating deals of $1,100 per square foot, yet
its overall revenue is not even close to what Wal-Mart appreciates.
Then, Amazon is regularly observed as the greatest danger to the Wal-
Mart plan of action. The web based business pioneer has been developing
deals by about 20% every year, creating upper hands en route and set to top
$100 billion in income this year. Incidentally, its prosperity has come from
88 Advances in Data Science and Computing Technology

various perspectives by taking pages from the Walton playbook – putting the
client first and offering the most reduced costs.

FIGURE 7.2 Walmart’s net sales worldwide from 2006 to 2019.

Source: statista.com.

7.3.2 IMAGINING THE WAL-MART OF THE FUTURE

Wal-Mart deals have eased back as of late as U.S. superstores have all the
earmarks of being arriving at an immersion point. In the course of the most
recent 12 years, the organization had the option to dramatically increase
income to its present sign of $488 billion, however, it will probably take any
longer for it to twofold by and by to about $1 trillion. Store extension has
additionally eased back altogether – all out U.S. supercenters expanded by
just 3% a year ago, and a lot of that expansion originated from changing over
rebate stores. Development was likewise delayed for global markets and
Sam’s Clubs. Conversely, the quantity of supercenters developed by 17% in
2005. Presently, the board is concentrating on two development roads: first is
its neighborhood Markets. In financial 2015, the organization included 232
of these littler configuration stores, developing the base by over half. Execu­
tion at these stores has been solid with same-store deals bouncing 7.7% in
the latest quarter. The subsequent open door is online business. Wal-Mart has
unmistakably fallen behind Amazon in online retail, yet it needs to bend over
Impact of Big Data on E-Commerce Websites 89

backward to make up for lost time. The organization acquired $12 billion in
online deals a year ago and expects to significantly increase that figure to
$35 billion by 2018 with interests in circulation focuses and innovation. All
things considered, contrasted, and almost $500 billion in complete income,
that scarcely moves the needle.

7.4 OBJECTIVE OF THE STUDY

Marketers analyze Big Data to match with the expectations of customers and
hence frame new product line and extend their product width. To streamline
the production and manufacturing performance, marketers require huge data.
But to assess the concept of Big Data from the customers perception is also
one of the huge factors that should be taken into account under this scenario.
Hence the objectives of the research could be categorized as under:
1. Primary Objective: The main objective of the chapter is to examine
the perception of E-Commerce websites towards big data and person­
alized campaigns and analyzing its impact on buying decision.
2. Secondary Objective: The research chapter also tried to analyze the
pros and cons regarding the recent big data analytics amalgamation.

7.5 ANALYSIS AND DISCUSSIONS

The business case for Big Data has been clear for quite a while, anyway over
the latest couple of years, we have seen critical strolls in how associations are
putting gigantic data to use to all the more probable fathom their own strategies
and how they convey and offer to customers. From a B2B standpoint, immense
data has quite recently exhibited to be a primary purpose in the movement
of online strategic policies and stages. It is prepared for associations to not
simply grasp the gigantic proportions of data they are gathering – yet likewise
how to all the more promptly understand what their customers are expecting
to get away from the modernized channels they are building.

7.5.1 EXPANDED EXAMINATION AND REVELATION OF

NOTEWORTHY BITS OF KNOWLEDGE

Colossal data gets past the overall wreck of customer data that associations
are gathering, zeroing in on the critical pieces of information that notice
90 Advances in Data Science and Computing Technology

to associations what they need to think about spikes famous, customer

tendencies, changes in buyer direct and that is just a glimpse of something
larger. Understanding buyer lead and their specific tendencies when making
a purchase are fundamental for continued with bargains and all in all
accomplishment.

FIGURE 7.3 Big data benefits in ecommerce.

Source: forbes.com.

An association probably will not have a particular commitment that

customers are searching for in their own thing set, yet there may be acces­
sories inside their organic framework that have the significant applications
and extensions that do. Colossal data empowers associations to fathom
what they need to take any action on, considering what customers need and
are looking for. Right when inspected dynamically, the move can be made
immediately, which can be the differentiation that addresses the decision
time an arrangement.

7.5.2 IMPROVED CLIENT ENCOUNTERS AND PURCHASER

VENTURES

In a U.S. study done by Help Scout, it was found that most of Americans
have given up a masterminded purchase on account of helpless help. This
Impact of Big Data on E-Commerce Websites 91

identical examination furthermore reported that associations lose more than

$62 billion consistently taking into account helpless customer backing and
experience. Clearly, customers are looking for positive experiences and bound
together buying adventures that eliminate the scouring from purchasing
on the web – coordinating this will be especially critical to a business’
fundamental concern. Customers that can find what they need, when they
need it are happy customers who become reiterate ones – especially when
the things they need are no matter how you look at it place with a clear a path
to purchase.

7.5.3 SIMPLER AND PROGRESSIVELY SECURE APPROACHES TO

FIND AND PAY ON THE WEB

Associations can use colossal data for a load of improvements in how

customers pay on the web and how these trades stay cautious. This separated
deception for constant similarly as tax avoidance deceives that appear as
though they are true trades. Since tremendous data consolidates unmistak­
able portion limits into one central stage it diminishes the threat of deception
while moreover making trades less complex for customers.
Joining and packaging data for logically direct, quick, and brilliant
assessment, tremendous data has prepared for associations to build up
their customer base while furthermore scaling express procedures, like
web business, for extended triumphs. As both the business and purchaser
universes continue making more data, it is inflexibly huge for associations
to stay before how they track, dismember, and put this data to authentic
use. For B2B web business to continue creating, it will require moving past
essentially looking at the data open–forefront advancements and assessment
approaches ought to be shipped off totally get the prizes and reinforce long
stretch accomplishment.
All things considered, the imbuement of tremendous data has given
associations an intuitive strategy to locate the critical encounters from
inside their instructive assortments that will advance business endeavors, for
instance, online business frameworks, progressively productive. Exploiting
huge data gives associations a lively and smoothed out a way to deal with
focus on significant pieces of data–which would then have the option to
be used to improve customer experience over the entire buying adventure,
making brisk proposals and unraveling approaches to purchase, while
keeping trades logically secure.
92 Advances in Data Science and Computing Technology

7.6 RECOMMENDATIONS

The research work is based to break down the significance of Big Data
for promoting of customized items which has just been restricted with the
respondents of Kolkata. The information for this exploration work just
centered around not many of the areas, for example, video spilling applica­
tions, and Amazon for instance to site the information. The territory for
customized promoting is very wide and need exceptional consideration
as large information bolsters in surrounding the legitimate prerequisites
of the clients in both on the web and disconnected markets. The part for
this research is very tremendous and can be beseeched through different
elements of focusing on and situating procedures utilized by advertisers
to investigate the customized showcases through large information. The
socioeconomics utilized right now be additionally investigated while the
locational portion could be expanded for a superior extent of research and
more extensive research space.

7.7 CONCLUSION

Personalization is undoubtedly an appealing vision for advertisers. A dream

which is at some point tempered by unreasonable desires, wrong application
approaches and security stricken laws that give clients more control on close
to home requests. Huge information without a doubt assumes a significant
job in alluring the vision of personalization and produce crusades remem­
bering potential clients. It gives a feeling that brands care for the clients and
henceforth clients pay off by indicating more dependability towards those
brands. According to a review directed by Forbes, 64% of the respondents
accept that information-driven advertising is fundamental for the develop­
ment of the current hyper-serious economy. This exploration concentrated
basically on to comprehend the client’s recognition of the customized
market and its significance through the investigation of large information.
The investigation demonstrated that however, individuals have less thought
on enormous information. Research yet firmly concurs that it is significant
for advertisers to infer ends and comprehend the purchasing conduct of the
clients, consequently concentrating on personalization. Respondents have
reasonable information about Marketing, information-driven markets, and
data innovation yet subliminally they are a piece of huge information and
consequently follow an extraordinary example that has been offered by the
Impact of Big Data on E-Commerce Websites 93

advertisers. The instances of Walmart and Amazon unmistakably depicted

that clients do acknowledge items which are customized by the advertisers for
explicit clients at certain purpose of time. In spite of the fact that an ongoing
McKinsey study of senior promoting pioneers expressed that solitary 15%
of the CMOs accept that their organization has been on a correct track with
personalization, however, today’s pioneers have discovered demonstrated
that there has been an expansion of 10–30% increment in deals. Henceforth,
one might say that personalization is not only a strategy to assemble brands;
it additionally presents upsell open doors for increment of deals.

KEYWORDS

• analytics for E-commerce

• big data analytics
• decision making
• E-commerce

REFERENCES

Data Analytics in E-Commerce Retail. https://towardsdatascience.com/data-analytics-in-e­

commerce-retail-7ea42b561c2f (accessed on 08 December 2021).
Ecommerce Insights on the Go. https://www.bigcommerce.com/blog/ecommerce-big-data/
(accessed on 08 December 2021).
How Big Data Analysis Helped Increase Walmarts Sales Turnover. https://www.dezyre.com/
article/how-big-data-analysis-helped-increase-walmarts-sales-turnover/109 (accessed on
08 December 2021).
How Big Data Analytics Has Changed Ecommerce Industry. https://www.smartdatacollective.
com/how-big-data-analytics-has-changed-ecommerce-industry/ (accessed on 08 December
2021).
How Big Data Benefits E-Commerce. https://dg1.com/blog/how-big-data-benefits-e-commerce
(accessed on 08 December 2021).
Three Key Benefits of Using Big Data in B2B E-Commerce. https://www.itproportal.com/
features/three-key-benefits-of-using-big-data-in-b2b-e-commerce/ (accessed on 08 December
2021).
Walmart Sales Forecasting. https://medium.com/analytics-vidhya/walmart-sales-forecasting­
d6bd537e4904 (accessed on 08 December 2021).
CHAPTER 8

A REVIEW OF THE REVOLUTIONIZING

ROLE OF BIG DATA IN RETAIL
INDUSTRY
MAUSUMI DAS NATH and MADHU AGARWAL AGNIHOTRI
Department of Commerce, St. Xavier’s College (Autonomous), Kolkata,
West Bengal, India, E-mails: [email protected] (M. D. Nath),
[email protected] (M. A. Agnihotri)

ABSTRACT

Information is present all around us in different kinds of forms. Voluminous

amount of information is generated daily in several forms. It has become
highly appreciable as information provides benefit both to the individuals
as well as to the business organizations. The existence of any business in
this competitive market depends solely on acquiring accurate information at
a particular time. Companies must have prior information concerning their
customers, products or inventory, market, and themselves too. To carry out
this task the structured and the unstructured data must be combined efficiently
to improve the business transactions, analyze the past performance, and
forecast the market advantages. This led to a new concept termed as Big
Data which companies rely on. There are several factors that contributed
to the exponential usage of today’s big data. It helps them to store, process,
and analyze data, monitor, and improve their performance and also generate
revenue. It helps to reengineer business models and also transform the
decision-making process. In this chapter, we portray some features of Big
Data and how it revolutionizes the decision-making process and the overall
performance of the business organization.
96 Advances in Data Science and Computing Technology

8.1 INTRODUCTION

Information is scattered everywhere in the environment and it is generated daily

through machines also. It is in the form of either unstructured, semi-structured
or structured form. It is to be captured, stored, processed, and analyzed for
variety of activities and events that occur all around. Thus, the concept of Big
data has emerged. In the last decade, giant companies like Google, Facebook,
and Amazon are all using this new technology to sustain in the global market by
gaining a competitive advantage over other organizations. Big data details about
events, preferences, and behaviors taking place all around and finally gives
access to a wider reach to the voluminous amount of data from gigantic and
multiple resources in a lesser amount of time. Companies collect and store data,
and then analyze to find out new revenue streams. The data are sorted so as to
process them in a lesser time. Different companies take the help of big data for
a reason of their own. For example, telecommunications company use big data
to assess the data traffic and the volume of incoming data. Amusement park like
Disney Land introduced “magic wristbands” to improve the attention of park
visitors. Prediction is carried out by wholesale company Amazon which helps
them to outline the customer’s behavioral pattern and their buying preferences.
The opportunities of big data is enormous. Companies can get a clear picture of
their future business value and thus big data helps in making a better decision
in their business activities. Big data finds its application in many areas like
telecommunications, IT sector, retail industry, finance, health as well as crime
to name a few. Hence, Big data has drawn up as a significant area of research
for many researchers, it finds its enormous impact in business environment.
Although it has several challenging issues, the benefits have solved many areas
of concern in many sectors of business. The implementation of big data gives an
efficient way to manage huge sets of data effectively in a cost effective manner
in lesser amount of time. Thus, big data helps in monitoring the performance of
the organization and ultimately enables in taking better decisions.
The presentation of the chapter is organized as follows. Section 8.2 focuses
on the literature review and Section 8.3 states the objectives of the study. Role
of big data in the retail sector is given in Section 8.4. Discussions are given in
Section 8.5. Section 8.6 illustrates the future directions and concluding remarks.

8.2 LITERATURE REVIEW

Big data finds its applications and advantages in a business environment.

Ajah and Nweke (2019) reviewed and discussed the upcoming trends,
A Review of the Revolutionizing Role of Big Data 97

business opportunities and limitations of big data and how it has helped the
organizations to create successful business strategies and remain competitive
in the business market. Furthermore, this chapter also brought forth the
different applications of big data and business analytics, how the data sources
were being generated in these applications and their main features. The
reviewers portrayed the challenges for a big data project and showed a path for
future research too. Again, Berntzen, and Krumova in 2017 applied Porter’s
value chain to big data and a framework has been suggested to figure out
the possible business opportunities for the business organization. The most
frequent revenue models have been discussed with its application. Moreover,
Banica and Alina focused (2015) on how big data helps organizations to
gain a competitive advantage over others. Moreover, it highlights how big
data helps to build, organize, and perform analysis from huge datasets by
providing a three-layered architecture. Furthermore, Pugna et al. in their
chapter (2019), outlined that there are some organizational challenges due to
the use of Big Data, but its impact on the business environment is enormous,
especially on performance management. Managers’ views and understanding
have been analyzed in terms of their performance in the company and also
on the decision-making process. After adopting the grounded theory, it has
been found that the key areas need special attention and understanding.
Gao, Chunli, and Chuanqi (2016) have discussed the validation and quality
assurance of big data, including the main concepts. Furthermore, the chapter
presented a comparison among big data validation tools and several major
market players were considered and discussed. Moreover, the primary
issues, challenges, and needs of big data have been discussed immensely. On
the other hand, Furht, and Flavio explained in the chapter (Furht and Flavio,
2016) the term big data and its role. It also highlights the main technological
aspects in a big data environment. Alsghaier et al. in their research (2017)
focused on some aspects of Big Data and its effect on organizations’ business
performance and how firms use the famous open source platform Hadoop to
process the huge amount of data. Alam et al. in their chapter (2014) focused
on the role of big data in the business and its various challenges. Why these
challenges are not dealt in the planning stage by the organizations have also
been discussed in details. Here, Alsghaier et al. (2017) illustrated on some
important points on Big data and its need on business performance and how
Hadoop can be used to process the data to win over others in this competitive
market. Again, Kubina, Michal, and Irena (2015) showed the path as to
how to reach a competitive advantage by using big data. Several research
techniques like content analysis, documents study, comparative analysis,
process analysis, statistical analysis, empirical research were used.
98 Advances in Data Science and Computing Technology

8.3 OBJECTIVES OF THE STUDY

The objective of this study is to explore the ways in which big data has played
a pivotal role in business. this study focuses on finding out the revolutionary
contribution of big data in retail sector. It is based on secondary data.

8.4 ROLE OF BIG DATA IN RETAIL

Various use cases of retails sector has been studied to understand the business
model of traditional retail sector and the retail sector where big data concept
has been implemented. The observations made are then analyzed to find out
the significant contributions of the technique adopted.

8.4.1 TRADITIONAL RETAIL BUSINESS MODEL

Retail business focuses mainly on delivering good shopping experience to

customers to retain them. Customer’s shopping experience is influenced by
various factors such as quality of the product or the service on offer, the channels
of delivery, time-frame, communication experience, payment flexibility,
incentives on offer, etc. Retail Business comprises of significant dimensions
such as customer value proposition, revenue model and operating model.
Customer value proposition is centered on the overall customer wellbeing. It is
about giving a good experience to customer so that they stay with the company
in long run. Customer’s satisfaction would be highest if the product or services
availed from a company are of utmost quality within reasonable price range.
Also, the way the facilities, services, and incentives are offered plays a critical
role in achieving higher customer satisfaction. Revenue model for business is
also crucial for business as the survival of business depends on customer as
well as the revenue line. Continuous flow of revenue is very important and
hence, business needs to have a systematic cost model and corresponding
earning model in place so that business can sustain in long run. The source
of money for business, how the money will be distributed through various
retail locations to meet up all expenses and target revenue to be generated
for smooth conduct of business are of critical importance. The distribution
of product and services on offer for customer, flow of funds (cost as well as
earning), and customer satisfaction are the key components that builds a sound
retail business system. The soundness of the system entirely depends on the
availability and accessibility to right data at right time (Figure 8.1).
A Review of the Revolutionizing Role of Big Data 99

FIGURE 8.1 Retail business model portraying two significant dimensions.

In traditional mode, the data management may be manual or digitized.

However, digitization does not always mean integrated systems. Digitized
system usually maintains data is maintained in flat file format. In such cases,
the issues of data replication, issues of multiple versions of same data, data
accessibility at various locations, maintenance of user profiles for filtered
data access, data status at various locations are serious concerns. Tradi­
tional retail business needs to transform their model to integrate customer
requirements, inventory, distribution channel activities, and build optimized
business model for smooth operations. The modern retail business model is
data driven model and hence, Big Data is playing a crucial role in providing
upgraded and integrated system.

8.4.2 RETAIL BUSINESS MODEL IMPLEMENTING BIG DATA

Due to ubiquitous nature of digital technology, customers can take informed

decisions about the products they want to buy or the services they want to
avail. As an amazing pool of updated information has become very handy for
online users, it has become very easy to compare the offers made by various
companies, price comparisons, reviews given by existing customers, and the
list is endless. Therefore, it is very essential for business to adapt technologies
that can provide deep insights into data to understand customer preferences
for the products, the manner in which those are offered, preferred delivery
100 Advances in Data Science and Computing Technology

channels, etc. Also, companies need to make strategies so that market capture
can be wide spread along with high profit.
Big data analytics is now being used in all retail business to identify
trends of existing scenario, forecast future demand, and determine pricing
strategies (https://www.forbes.com/sites/bernardmarr/2015/11/10/big-data­
a-game-changer-in-the-retail-sector/#356bf9099f37). The benefits of Big
data in the retail industry is huge which is illustrated by Figure 8.2.

FIGURE 8.2 Benefits of big data in the retail industry.

1. Trend Identification: This is achieved on the basis of the historical

data available with the company about customers. Existing pattern
of customers can help to find out the purchase pattern of customers
throughout the year. Also, region-wise list as well as season-wise
trends can also be identified with in depth analysis. Also, special
machine learning algorithm can be developed to filter out the data
available on social media, communities, forums, and blogs to
understand the interest of potential customers. It helps to understand,
the purchase preference of existing customer. Browsing habit of
people also tells a lot about their thought process which may be highly
affected due to hype in social media or in non-digital platforms also.
Retailers must deploy tools to drag out such minute details about
customers to read their mind for strategic planning.
2. Predicting Trends: Nowadays, retailers use a variety of tools
available to them to find out the season’s trends. They can choose
A Review of the Revolutionizing Role of Big Data 101

what items they should have in their shelves so as to attract the

customers, both old and new. Trend forecasting algorithms extract
the information from social media websites and web browsing
history to analyze out what is causing a murmur, and ad-buying
data is analyzed to find out what marketing departments will be
focusing on.
3. Demand Forecasting: Once trends have been identified, through
Machine Learning tool, future demand to be raised by customers for
the products and the services can be predicted precisely. Demographic
detail of customer reflects lot of information. Country’s economic
condition and the related parameters also help to understand the
scenario of the market and accordingly gives a good idea about
future demand. Market Basket Analysis helps retailers to identify the
customer preferences very clearly. Proper future assessment helps
retailer to assess the requirements well in advance and accordingly
enables them to tap the opportunity of high sales in a specific season
or region.
4. Price Optimization: Trend and demand helps to understand
clearly the market scenario and customer response. Accordingly, to
tap the potential customers and to retain the existing ones, retailer
needs to provide products and services in the right combination of
price and quality keeping in mind several factors such as season,
region, competitor’s offerings, economic indicators. Pricing
strategies plays a crucial role in the survival of the business. Big
data analytics helps business to understand the price scenario and
determine the price fluctuation scenario for their own company. It
can give clear indications for the right time for price hike or price
drop.
5. Market Identification: Big data analytics helps tremendously in
finding out which customers will be buying which type of product in
a specified time period in a specified region. This helps the retailer
to be ready with the required stocks, as demand has been assessed
accurately. It also enables retailers to be ready with the delivery
channel and all the logistics in place so that timely order fulfillment
can be achieved.
6. Market Segmentation: Big data enable the retail industry to segment
the market. They narrow down their segments as “young minds,’
“digital people,” Senior citizens,” etc. Then they again sub categorize
as sport lovers, movie viewers, etc. Based on these categories and
102 Advances in Data Science and Computing Technology

sub-categories, the success or the failure rate depends (https://indata­

labs.com/blog/impact-of-big-data-on-business). In the traditional
retail business model, marketing was mainly a one-way communica­
tion. Instant Customer communication or feedback was not prevalent.
But, with the advent of big data and application of various algorithms,
market segmentation has become more easy and effective. One-way
communication have been transformed into a bi-directional commu­
nication. Applications of this concept of big data can remove certain
segmentation barriers and help the retail industry on a micro level with
personalized, customized messages to which the potential customers
can relate and identify.
7. Payment Patterns: It has been observed that even though many
online visitors visits the retailer website and add product to their
cart, some do not complete the final payment process due to various
reasons. Data analytics has reflected that the hesitation in making
online payment, not being convinced about the safety measures on
the retailer website, required card is not available with the customer at
that moment, inability to complete the online transaction within valid
session time frame and few more such reasons prohibits the ultimate
conversion of the visitor into customer (https://www.simplilearn.
com/big-data-transforming-retail-industry-article). Thus, multiple
payment mode facility must be deployed on the retailer website for
having high conversion rate of visitors into customers.
8. Informative Dashboards: Highly informative and attractive dash­
boards on retailer website puts a very good impression on the minds
of customers. The matrices about the product performance, market
penetration, customer feedback, can be accurately calculated and put
up on the dashboard with the help of analytics tools.
9. Making Strategic Decisions: Companies are highly benefitted by
the enormous amount of information extracted, gathered, stored, and
analyzed. The analysis helps them to take a strategic marketing deci­
sion in all their activities in future.
10. Sensitivity Analysis: It is a mathematical model which marketers
often use to find out the uncertainty in the market. By applying
sophisticated machine learning based algorithms, companies can
quantify the uncertainties prevailing in the market and figure out the
optimal parameter settings required in their model. This data can be
used to accurately predict what the top selling products or brands in
a category would be.
A Review of the Revolutionizing Role of Big Data 103

8.5 ANALYSIS AND DISCUSSIONS

Modern retail business can be highly benefitted with the utilities big data
can bring into. Retail business has always been working with lot of data
to understand customer scenario and accordingly designing the strategies
to tap the market. However, big data analytics helps them to get hold of
unstructured data also so that better understanding of customers can be
gained. This ultimately helps in managerial as well as operational efficiency
in retail sector.
As compared to the traditional retail model, big data help companies
aim at offering improved customer services, better product, and pricing
services, which thereby can help to quantify their profit. As retail sector
focuses primarily on customers, big data enables the companies to retain
their old customers and extend their customer base too. Other facts include
better target marketing, cost reduction, and improved efficiency of existing
processes which leads to better performance and helps them to sustain in this
competitive market.
Thus, Big data help business organizations to analyze information and
improve on taking better and effecting strategic decisions. Moreover, the retail
sector collects a large amount of data through RFID, POS scanners (https://
indatalabs.com/blog/impact-of-big-data-on-business), various customer loyalty
programs, etc. Application of big data enables in detecting and reducing frauds
and enables the timely analysis of inventory. Earlier, big data was mainly
implemented by large business houses that could afford the technologies but,
recently, both large and small business enterprises are highly relying on this
game changing concept of big data for intelligent business purposes. Hence,
the demand for big data proliferates in the market domain and specifically in
the retail segment. Enterprises from across different industries consider ways of
how big data can be used in business. Therefore, the benefit of big data are to
improve productivity, identify customer needs and demands, offer a competitive
advantage, and chalks out ways for sustainable economic development. Even
business owners are increasingly investing in big data solutions to optimize
their operations and manage data traffic. Vendors are adopting big data solu­
tions for better procurement methods, hassle-free supply chain management,
etc., to work on data efficiently and effectively. The integration of the different
business processes has enabled the strategic managers to take decisions accu­
rately and timely. The benefits and its impact is so huge in the retail industry,
that companies who had not implemented this concept are thinking of shifting
to this technological advancement (Figures 8.3 and 8.4).
104 Advances in Data Science and Computing Technology

FIGURE 8.3 Big data scenario (in %) in 2018.

Source: https://indatalabs.com/blog/impact-of-big-data-on-business.

FIGURE 8.4 Rising trend of big data market by 2025.

Source: https://indatalabs.com/blog/impact-of-big-data-on-business.

8.6 CONCLUSION AND THE FUTURE SCOPE

From Figure 8.4, it is seen that the big data market is expected to witness
remarkable growth by the year 2025. An important reason is an exponential
growth in the volume of both structured and unstructured data. Apart from
other factors increased technological penetration in the market and the
A Review of the Revolutionizing Role of Big Data 105

wide usage of smart phones has led to the generation of larger amounts
of data. The escalating need for analyzing data will lead to the rise of
demand for big data over the forecast period. Furthermore, the number of
online businesses in the industry is also growing, owing to enhanced profit
margins. Other industries, such as healthcare, utilities, oil, and gas, logistics,
manufacturing, and banking, will widely use online platforms to provide
customized d services to customers. These have led to the rise of global big
data market growth. Even, socio-economic benefits are also associated with
big data. Therefore, several government agencies have laid down policies
for promoting the development of big data. But, usage of big data in several
industries, such as healthcare, oil, and gas, and so on, has been growing in a
snail’s speed. Adopting new technology, implementing certain sophisticated
algorithms based on data analytics, AI, and quantum computing has been a
challenging task. Moreover, it is expensive to adopt and integrating different
operational units is cumbersome. Furthermore, there should be strong and
efficient security measures to combat the data breaches if it occurs.

KEYWORDS

• big data
• machine learning tool
• retail industry
• strategic decision-making

REFERENCES

Ajah, I. A., & Henry, F. N., (2019). Big data and business analytics: Trends, platforms, success
factors and applications. Big Data and Cognitive Computing, 3(2), 32.
Alam, J. R., Asma, S., Ramzan, T., & Muneeb, N., (2014). A review on the role of big data in
business. International Journal of Computer Science and Mobile Computing, 3(4), 446–453.
Alsghaier, H., Mohammed, A., Issa, S., & Samah, A., (2017). The impact of big data analytics on
business competitiveness. Proceedings of the New Trends in Information Technology (NTIT).
Alsghaier, H., Mohammed, A., Issa, S., & Samah, A., (2017). The importance of Big Data
Analytics in business: A Case study. American Journal of Software Engineering and
Applications, 6(4), 111–115.
Banica, L., & Alina, H., (2015). Big data in business environment. Scientific Bulletin-Economic
Sciences, 14(1), 79–86.
106 Advances in Data Science and Computing Technology

Berntzen, L., & Krumova, M., (2017). Big Data from a Business Perspective, Conference
Paper. doi: 10.1007/978-3-319-65930-5_10.
Furht, B., & Flavio, V., (2016). Introduction to big data. In: Big Data Technologies and
Applications (pp. 3–11). Springer, Cham.
Gao, J., Chunli, X., & Chuanqi, T., (2016). Big data validation and quality assurance--issuses,
challenges, and needs. In: 2016 IEEE Symposium on Service-Oriented System Engineering
(SOSE) (pp. 433–441). IEEE.
https://indatalabs.com/blog/impact-of-big-data-on-business (accessed on 08 December 2021).
https://www.digimarc.com/resources/retail-tech-spendingreport?utm_source=google&utm_
campaign=planet%20retail&utm_medium=ppc&gclid=EAIaIQobChMIosXP88Pz5wIVS
A4rCh3TxAP-EAAYASAAEgLsRPD_BwE (accessed on 08 December 2021).
https://www.forbes.com/sites/bernardmarr/2015/11/10/big-data-a-game-changer-in-the­
retail-sector/#356bf9099f37 (accessed on 08 December 2021).
https://www.simplilearn.com/big-data-transforming-retail-industry-article (accessed on 08
December 2021).
Kubina, M., Michal, V., & Irena, K., (2015). Use of big data for competitive advantage of
company. Procedia Economics and Finance, 26, 561–565.
Pugna, I. B., Adriana, D., & Oana, G. S., (2019). Corporate attitudes towards Big Data and its
impact on performance management: A qualitative study. Sustainability, 11(3), 684.
 PART III

Algorithm for Load Balancing

in Cloud Computing

CHAPTER 9

LOAD-BALANCING IMPLEMENTATION
WITH AN ALGORITHM OF MATRIX
BASED ON THE ENTIRE JOBS
PAYEL RAY, ENAKSHMI NANDI, RANJAN KUMAR MONDAL, and
DEBABRATA SARDDAR
Department of Computer Science and Engineering, University of
Kalyani, West Bengal, India, E-mails: [email protected] (P. Ray),
[email protected] (E. Nandi), [email protected]
(R. K. Mondal), [email protected] (D. Sarddar)

ABSTRACT

Cloud computing is the on-demand delivery of IT resources over the internet

with pay-as-you-go pricing. Instead of purchasing, retaining, and managing
personal data centers and servers, you can access technology services, such as
computing capability, storage, and databases, on as-necessary support from
a cloud provider. Cloud computing is internet-based computing. There are
a more significant number of servers associated with the system to provide
many kinds of network services to give cloud clients. Deficient quantities
of servers related to the cloud have to produce extra per task at a time. So,
it is a complex issue to perform all tasks at a peculiar time. Some systems
carry out all tasks, so there is a need to readjust all amount of loads at a time.
Load balancing adjusts the closure time and presents all tasks, respectively.
There is not available always to remain a corresponding amount of servers
to carry out the same tasks. Tasks to be dealt with in cloud computing would
be higher than the linked servers. Inadequate servers have to go on a million
volumes of tasks at a time. We propose algorithms where some machines
110 Advances in Data Science and Computing Technology

perform the jobs here. The jobs are higher than the machines and balance all
machines to exploit the quality of services in cloud computing.

9.1 INTRODUCTION

Cloud system is a just now progressing procedure to give online sources,

storage, and enable consumers to manage applications with improved
scalability, ease of use, and fault tolerance (Chan, Jiang, and Huang, 2012).
The Cloud system is about storing the substance on isolated workstations
instead of on individual physical systems or new machines (Jeevitha,
Chandrasekar, and Karthik, 2015). This instruction can turn back accepting
the network on any machine, quite where in the world, if that machine can
support cloud system systems (Chen et al., 2013). The cloud system contains
a front-end, that is the user side, and a back-end that is an association of
the workstations and physical machines held by an agent keeping the data
(Krakowiak, 2009). A central workstation that is a splinter of the back-end
following protocols and uses middleware to be in contact between networked
physical systems (Dinh, Lee, Niyato, and Wang, 2013). The Cloud system
collects all the computing resources and deals with them mechanically
(Pareek, 2013). Its features characterize a cloud system: on-need self-
service, pooling of sources, perfect entry to the network, the flexibility of
service convenience, and efficiency of services utilized by individual clients
(Mansouri and Buyya, 2016; Sibiya, Venter, and Fogwill, 2015). Cloud
system has hugely been with devices like Google Drive replacing Microsoft
Office, Amazon Web Services re-establish traditional business data storage,
websites replacing branch agencies, and Dropbox storing all our data and
files (Kabakus and Kara, 2017).

9.2 CLOUD COMPUTING FEATURES

Cloud computing features are as following (Kaur and Luthra, 2012):

1. Service Provides on-Demand: When any applicant demands services
and resources, thus the cloud deliver services on demand.
2. Rapid Flexibility: The number of resources in the cloud can be raised
and deteriorate smoothly.
3. Resource Pooling: Resources are designated at different positions
corresponding to the buyer’s demands.
Load-Balancing Implementation with an Algorithm 111

4. Pay per Use: Conforming to the consumer’s application of computing

resources is asked to be paid off.

9.3 LOAD BALANCING

In the cloud computing background, load-balancing is a method to share

workloads amongst their multiple machines to avoid overload problems
(Ray and Sarddar, 2018). It assists in controlling the resources and
services (Chaudhari and Kapadia, 2013). It makes better utilization of
resources and assists in improving the act of the system. Load balancing
is the technique to determine the overloaded machine (cloud worksta­
tions) and help transfer the extra load to other machines. The following
Figure 9.2 step is as:
1. First, a client sends a demand for service or resource;
2. Then though, web demand goes to load balancing (CLB);
3. From CLB, the demand goes to the Database workstation through
a cloud workstation. After that Database workstation full field, the
demand in a proper manner;
4. In this way, incoming, and outgoing demand and the reply will
continue to carry out with any failure. It divides load balancing
algorithms into two groups: static load balancing and dynamic load
balancing algorithm:
i. Static Load Balancing Algorithms: The loads do not depend
on the system’s present status, but it needs facts concerning the
nodes’ uniqueness. This type of algorithm partitions the traffic
uniform among the workstations. By this approach, the traffic
on the workstations will be handling simple, and consequently,
it will make the situation better.
ii. Dynamic Load Balancing Algorithms: Dynamic algorithms
are more flexible than the static algorithms, and they do not
rely on prior learning but depends on the present status of the
system. A dynamic algorithms algorithm search throughout
the whole network and selects the appropriate weights on the
workstation, and it refers to the lightest workstation to balance
the traffic. However, selecting a suitable workstation requires
secure communication within networks, leading to better traffic
attached to the system.
112 Advances in Data Science and Computing Technology

9.4 RELATED WORKS

Cloud computing presents a variety of utilities to the end-user, such as media

sharing, online software, gaming, and online storage. In a cloud environment,
all nodes represent all tasks or subtasks (Ritchie and Levine, 2005). The
opportunistic load balancing algorithm (OLB) plans to stay all nodes
active despite the current workloads of all nodes (Braun et al., 2001). The
OLB algorithm allocates tasks to produce nodes randomly. The minimum
completion time (MCT) algorithm allows all tasks to the nodes, considering
the scheduled MCT of this task over other nodes (Wang et al., 2010).
The Min-Min scheduling algorithm (MM) considers the same scheduling
approach as the MCT algorithm to allot each task, the node to end this task
with MCT over other nodes (Che-Lun, Hsiao-His, and Yu-Chen, 2012). The
load balance Min-Min (LBMM) algorithm (Kokilavani and George, 2011)
enforces the MM scheduling technique and load balancing program. It can
escape the unnecessarily duplicated assignment (Min-You, Wei, and Hong,
2000). Load Balancing with Job Switching (Ray and Sarddar, 2018) minimize
loads of oppressive loaded system to under loaded machine by replacing a
specific task way. Load Balancing of an Unbalanced Matrix with Hungarian
method (Ranjan et al., 2017) utilizes the Hungarian algorithm, where tasks
are more extensive than all systems. The load balancing of the unbalanced
cost matrix (Ranjan et al., 2017) is the same as the other algorithm.

9.5 PROPOSED METHOD

To determine the matrix cost as well as a combination of tasks(s) vs

machine(s) of an unbalanced matrix, to concentrate on a problem consisting
of ‘m’ machines M={M1, M2, …, Mm}. A set of ‘n’ jobs J={J1, J2, …, Jn}
is considered to be assigned for execution on the ‘m’ available machines
and the execution cost Cij, where i = 1, 2, …, m, and j = 1, 2, …, n are
mentioned in the cost matrix where m>n. First of all, we obtain the figure for
each row and each column of the matrix; keep the results in the array, i.e.,
Row-sum, and Column-sum. Then we select the first n rows by Row-sum,
i.e., starting with most minimums to next minimum to the array Row-sum
and deleting rows corresponding to the left behind (r) jobs. Accumulate the
results in the new array that should be the array for the first sub-problem.
Repeat this process until remaining jobs become less than a machine when
remaining jobs are less than n, then, deleting (c) columns by Column-sum,
Load-Balancing Implementation with an Algorithm 113

i.e., the corresponding value(s) maximum to the next maximum to form the
last sub-problem. Save the results in the new array that shall be the array for
the last sub-problem.

9.5.1 HUNGARIAN METHOD

Steps of Hungarian method (Kuhn, 1955) as follows:

 Step 1: The input of this algorithm is an n×n square.

 Step 2: Find the smallest element from each row and subtract it from each
element in the corresponding row.
 Step 3: Similarly, for each column, find the smallest element and subtract it
from each element in the corresponding column.
 Step 4: Cover all zeros in the subtracted matrix with a minimum number of
horizontal and vertical lines. If lines numbers are n, then an optimal assign­
ment exists. The algorithm stops. Otherwise, if lines numbers are less than n
then, go to the next step.
 Step 5: Get the smallest element that is not covered by a line in Step 4,
deduct with the lowest uncovered element from each uncovered element, and
include the lowest uncovered element to each element covered two times.

9.6 PROPOSED ALGORITHM

To present an algorithmic representation of the method, Consider a problem

that consists of a set of ‘m’ Machines M = {M1, M2, …, Mm}. A set of ‘n’ jobs
J = {J1, J2, …, Jn} is considered to be assigned for execution on ‘m’ available
machines and the execution cost Cij, where i = 1, 2, …, m, and j = 1, 2, …,
n, where m>n, i.e., the number of jobs is more than a number of machines.

 Step 1: m×n matrix.

 Step 2: Add all elements of each column of the matrix, namely, Column-sum
in each column, respectively.
 Step 3: Select the first k column by Column-sum, i.e., starting with most
minimums to next minimum to the array Column-sum until some rows and
number of columns are equal. Make a new balanced problem.
 Step 4: Concern Hungarian method to solve the problem.
 Step 5: List the element costs of identical machines.
 Step 6: Stop.
114 Advances in Data Science and Computing Technology

9.7 ILLUSTRATION OF AN EXAMPLE

Consider a problem in which a set of 5 machines M={M1, M2, M3, M4, M5},
and a set of 8 jobs J={J1, J2, J3, J4, J5, J6, J7, J8}. The assignment matrix
contains the execution costs of every job to each machine.

 Steps 1 to 2: Input: 5×8 matrix.

Jn/Mm M1 M2 M3 M4 M5
J1 151 277 185 276 321
J2 245 286 256 264 402
J3 246 245 412 423 257
J4 269 175 145 125 156
J5 421 178 185 425 235
J6 257 257 125 325 362
J7 159 268 412 256 286
J8 365 286 236 314 279

 Step 3: To obtain the sum of a column of the matrix, i.e., the sum of each
column is as follows:
Jn/Mm M1 M2 M3 M4 M5
J1 151 277 185 276 321
J2 245 286 256 264 402
J3 246 245 412 423 257
J4 269 175 145 125 156
J5 421 178 185 425 235
J6 257 257 125 325 362
J7 159 268 412 256 286
J8 365 286 236 314 279
Column-Sum 2,113 1,972 1,956 2,408 2,298

 Step 4: To equal rows and columns, we add some columns as dummy

based on minimum column-sum until a number of rows and number of
columns will be equal.
Load-Balancing Implementation with an Algorithm 115

Jn/Mm M1 M2 M3 M4 M5
J1 151 277 185 276 321
J2 245 286 256 264 402
J3 246 245 412 423 257
J4 269 175 145 125 156
J5 421 178 185 425 235
J6 257 257 125 325 362
J7 159 268 412 256 286
J8 365 286 236 314 279
Column-Sum 2,113 1,972 1,956 2,408 2,298

 Step 5: This corresponds to the following optimal cost in the original cost
matrix using Hungarian method.
Jn/Mm M1 M2 M3 M4 M5
J1 151 277 185 276 321
J2 245 286 256 264 402
J3 246 245 412 423 257
J4 269 175 145 125 156
J5 421 178 185 425 235
J6 257 257 125 325 362
J7 159 268 412 256 286
J8 365 286 236 314 279

 Step 6: Assign all jobs to corresponding machines.

Jn/Mm M1 M2 M3 M4 M5
Jn 310 423 361 264 156

Final Result
M1 → J1*J7 → 151 + 159 = 310
M2 → J3*J5 → 245 + 178 = 423
M3 → J6*J8 → 125 + 236 = 361
M4 → J2 → 264
M5 → J4 → 156
116 Advances in Data Science and Computing Technology

9.8 RESULT ANALYSIS

Figure 9.1 shows our proposed work result. Figure 9.1 explains the execution
time for each task at each node. To analyze the performance of our approach
is judged by other methods shown in Figure 9.1. Figure 9.1 displays the
comparison of the execution time of each computing node among our
approach. The total completion time for completing each task by using the
proposed work, LBMM, and MM, are 17, 27, and 42 ms, respectively. Our
approach attains the least completion time with enhanced load balancing
than available other algorithms.

FIGURE 9.1 Execution time (ms) of each task at each computing node.

9.9 CONCLUSION

The result of this problem, the matrix, can be solved with the support of
the Hungarian method. It gives each task or subtask with minimal comple­
tion time to its corresponding machine so that our result is excellent for
the system. Though some machines execute more than one task or subtask
completion time is excellent, and loads are being balanced with our proposed
work. Without some alternative consideration, our approach has continu­
ously been satisfied in various cases and prospects; we figure out a further
way of dealing with our difficulties sometimes.
Load-Balancing Implementation with an Algorithm 117

KEYWORDS

• load balancing
• minimum completion time
• unbalanced matrix

REFERENCES

Braun, T. D., Siegel, H. J., Beck, N., Bölöni, L. L., Maheswaran, M., Reuther, A. I.,
Robertson, J. P., et al., (2001). A comparison of eleven static heuristics for mapping a
class of independent tasks onto heterogeneous distributed computing systems. Journal of
Parallel and Distributed Computing, 61, 810–837.
Chan, M. C., Jiang, J. R., & Huang, S. T., (2012). Fault-tolerant and secure networked storage.
In: Digital Information Management (ICDIM), 2012 Seventh International Conference on
(pp. 186–191). IEEE.
Chaudhari, A., & Kapadia, A., (2013). Load balancing algorithm for azure virtualization with
specialized VM. Algorithms, 1, 2.
Che-Lun, H., Hsiao-His, W., & Yu-Chen, H., (2012). Efficient load balancing algorithm for
the cloud computing network. In: International Conference on Information Science and
Technology (IST 2012) (pp. 28–30).
Chen, C. H., Lin, H. F., Chang, H. C., Ho, P. H., & Lo, C. C., (2013). An analytical framework
of a deployment strategy for cloud computing services: A case study of academic websites.
Mathematical Problems in Engineering, 2013.
Dinh, H. T., Lee, C., Niyato, D., & Wang, P., (2013). A survey of mobile cloud computing:
Architecture, applications, and approaches. Wireless Communications and Mobile Computing,
13(18), 1587–1611.
Jeevitha, M., Chandrasekar, A., & Karthik, S., (2015). Survey on verification of storage
correctness in cloud computing. International Journal of Engineering and Computer Science,
4(09).
Kabakus, A. T., & Kara, R., (2017). A performance evaluation of drop box in the light of
personal cloud storage systems. International Journal of Computer Applications, 163(5).
Kaur, R., & Luthra, P., (2012). Load balancing in cloud computing. In: Proceedings of Inter­
national Conference on Recent Trends in Information, Telecommunication and Computing,
ITC.
Kokilavani, T., & George, A. D. I., (2011). Load balanced min-min algorithm for static meta­
task scheduling in grid computing. International Journal of Computer Applications, 20(2),
43–49.
Krakowiak, S., (2009). Distributed under a Creative Commons license. Middleware Architecture
with Patterns and Frameworks. http://creativecommons.org/licenses/by-nc-nd/3.0/ (accessed
on 10 January 2022).
118 Advances in Data Science and Computing Technology

Kuhn, H. W., (1955). The Hungarian method for the assignment problem. Naval Research
Logistics Quarterly, 2(1, 2), 83–97.
Mansouri, Y., & Buyya, R., (2016). To move or not to move: Cost optimization in dual cloud-
based storage architecture. Journal of Network and Computer Applications, 75, 223–235.
Min-You, W., Wei, S., & Hong, Z., (2000). Segmented min-min: A static mapping algorithm
for meta-tasks on heterogeneous computing systems. In: Heterogeneous Computing
Workshop, 2000, (HCW 2000) Proceedings, 9th (pp. 375–385). IEEE.
Pareek, P., (2013). Cloud computing security from single to multi clouds using secret sharing
algorithm. International Journal of Advanced Research in Computer Engineering &
Technology, 2(12), 3261–3264.
Ranjan, K. M., Payel, R., Enakshmi, N., Biswajit, B., Manas, K. S., & Debabrata, S., (2017).
Load balancing of unbalanced matrix with Hungarian method. In: International Conference
on Computational Intelligence, Communications, and Business Analytics (pp. 256–270).
Springer, Singapore.
Ranjan, K. M., Payel, R., Enakshmi, N., Priyajit, S., & Debabrata, S., (2017). Load balancing
of the unbalanced cost matrix in a cloud computing network. In: Computer, Communication
and Electrical Technology: Proceedings of the International Conference on Advancement of
Computer Communication and Electrical Technology (ACCET 2016) (p. 81). West Bengal,
India. CRC Press.
Ray, P., & Sarddar, D., (2018). Load balancing with inadequate machines in cloud computing
networks. In: Mandal, J., & Sinha, D., (eds.), Social Transformation – Digital Way. CSI 2018,
Communications in Computer and Information Science (Vol. 836). Springer, Singapore.
Ritchie, G., & Levine, J. A., (2005). Fast, effective local search for scheduling independent
jobs in heterogeneous computing environments. Journal of Computer Applications, 25,
1190–1192.
Sibiya, G., Venter, H. S., & Fogwill, T., (2015). Digital forensics in the Cloud: The state of the
art. In: IST-Africa Conference, 2015 (pp. 1–9). IEEE.
Wang, S. C., Yan, K. Q., Liao, W. P., & Wang, S. S., (2010). Towards a Load Balancing in a
three-level cloud computing network. In: CSIT (pp. 108–113).
CHAPTER 10

LOAD-BALANCING ALGORITHM
WITH OVERALL TASKS
ENAKSHMI NANDI, PAYEL RAY, RANJAN KUMAR MONDAL, and
DEBABRATA SARDDAR
Department of Computer Science and Engineering,

University of Kalyani, West Bengal, India,

E-mail: [email protected] (E. Nandi)

ABSTRACT

The innovative system cloud is a model that is based on virtualization. It

is a new name of existing technology. Cloud computing mainly serves
computing resources as a service to cloud clients. This new system is used
to store and access hard drives. This new computing paradigm provides
marvelous opportunities to solve the large-scale scientific issues. To utilize the
applications of cloud in different field, several challenges have to face, in that
case task scheduling is one of the important factor. In a cloud system, optimum
utilization of computing resources is always challenging. A cloud service giver
ensures to serve computing resources efficiently to cloud client at optimum
cost. Load balancing techniques play a vital role in the competent deployment
of computing entities. Different algorithm related to load balancing migrates
an overloaded virtual machine task to an under the loaded machine, with
disturbing the existing background. Various load balancing techniques are
suggested by cloud researchers. Hence, a resourceful load balancing process
has required. This research chapter depicts an assessment of the various loads
balancing method in the cloud system. Also, a comparative analysis has
been presented based on various performance measuring parameters. Cloud
120 Advances in Data Science and Computing Technology

computing is the convenient aptitude application on the Internetworking.

Cloud service providers deal with applications and resources. It faces different
challenges. Load balancing related problem is great issue. Load balancing is
nothing but proper distribution of the tasks among different nodes for the best
utilization of resources. The major target of a variety of type load balancing
algorithms is to decrease the waiting time and turnaround time. Within this
chapter, we are proposing a new scheduling technique in a distributed system
is choosing suitable entity with their total task. It is a very easy way to decide
on a suitable node. This idea would provide the effective usage of computing
entities and preserve cloud load balancing.

10.1 INTRODUCTION

Cloud computing is an example that interiors a collection of clients with

admittance to scalable, virtual nature of resources over the network (Chan,
Jiang, and Huang, 2012). Cloud computing is a form to extract and run
IT resources (Jeevitha, Chandrasekar, and Karthik, 2015). In data centers,
applications are run in physical workstations that are often provisioned with
a high workload (Chen et al., 2013). This configuration makes data centers
costly to maintain with significant overhead (Krakowiak, 2009). Data centers
are more flexible, secure, and give excellent support for on-demand resource
allocations (Dinh, Lee, Niyato, and Wang, 2013). It abstracts workstation
diversity, performs workstation consolidation, and enhances workstation
utilization (Pareek, 2013). A host is able to run multiple virtual machines with
various resource specifications and varied workloads (Mansouri and Buyya,
2016). Physical workstations hosting in homogeneous virtual machines with
unpredictable workloads may cause unbalanced resource usage (Sibiya,
Venter, and Fogwill, 2015).
Cloud data centers are extremely unpredictable due to (Kabakus and
Kara, 2017; Youseff, Butrico, and Da Silva, 2008; Jain and Paul, 2013):
• Irregular buyer demand pattern;
• Wavering resource usage by virtual machines;
• Various rates of arrivals and departure of customers; and
• The performance of the host may differ when handling various load
levels (Kumar and Lu, 2010). These reasons are enough to trigger
unbalanced loads in the data center, which needs a load balancing
mechanism to keep away from presentation degradation and service
level agreement destructions (Wang et al., 2010).
Load-Balancing Algorithm with Overall Tasks 121

10.2 CLOUD COMPUTING AND LOAD BALANCING

10.2.1 CLOUD SYSTEM

Cloud computing is an on-demand set of connections admittance to a team

of organize able entities that is able to be hastily prerequisite and unconfined
with nominal management endeavor (Moura and Hutchison, 2016). The
Cloud technology is based on virtualization concept which is the prime aspect
of this system. Virtual machine is a software completion of the physical entity
(Zhang et al., 2011). A hypervisor, which is called a virtual machine monitor,
is accountable for distribution a single physical occurrence of cloud entities
among diverse occupants (Sabahi, 2011).
The main aim of this computing is to progress the presentation of the
system and to get optimal resource exploitation, greatest throughput, highest
response time, and evading overload (Mishra and Mishra, 2015).

10.2.2 CLASSIFICATION OF LOAD BALANCING TECHNIQUE

Load balancing technique have been divided based on the present state of the
system and who instigated the process (Gupta and Deshpande, 2014).
1. Depending on which Client Initiates the Method:
i. Sender-Initiated: Sender or client begins the completion of the
load balancing technique helps to identify necessitate for load
balancing (Eager, Lazowska, and Zahorjan, 1985).
ii. Receiver-Initiated: Workstation initiates the completion of
the load balancing process to detect the requirement for load
balancing (Adhikari and Patil, 2012).
iii. Symmetric: This category of the algorithm is a blend of the
sender-beginner category and receiver-start up category methods
(Eager, Lazowska, and Zahorjan, 1985).
2. Depending on the Existing Static System Algorithm: In the
static algorithm, there is a uniform distribution of traffic among
the workstations (Vakkalanka, 2012). This algorithm needs a prior
understanding of system entities such as the judgment of shifting the
load does not rely on the running state of the system (Mani, Suresh,
and Kim, 2005). The static method is perfect in the system, which
has fewer inequalities in load (Kumar et al., 2001).
3. Dynamic Algorithm: In case of dynamic method for balancing
heavy load, the lightest workstation in the whole system is looked
122 Advances in Data Science and Computing Technology

upon and preferred (Patil and Shedge, 2013). For statement with the
set of connections is essential, which is able to enlarge the traffic in
the system (Akyildiz et al., 1999).

10.3 LOAD BALANCING

Load Balancing specifically signifies to transfer loads from overloaded

process to other under-loaded processes (Ansari and Sachin, 2015). The
main difficulty is how to choose the best node to execute a scrupulous job
and transfer a load to a suitable node (Bittencourt, Miyazawa, and Vignatti,
2012). Based on the implementation method, load balancing algorithms can
be divided in two ways (Willebeek-LeMair and Reeves, 1993):
• Static algorithms; and
• Dynamic algorithms.
Obtainable load balancing methods in cloud computing regard as a range
of parameters like (Rastogi, Bansal, and Hasteer, 2013; Bhargava and Goyal,
2013; Sran and Navdeep, n.d.):
1. Throughput: To estimate the number of jobs, those executions have
been accomplished.
2. Fault Tolerance: It signifies capability of an algorithm to achieve
consistent load balancing despite of random node or link letdown.
3. Migration Time: It is the time to migrate the task or entities from
one node to another node.
4. Response Time: The quantity of time required to reply to a meticu­
lous load balancing method in a scattered system.
5. Resource Utilization: It is the method to ensure the deployment of
resources.
6. Scalability: It is the metrics measures capability of an algorithm to
achieve load balancing for a scheme with some fixed number of nodes.
7. Performance: It is the process to test out the competence of the system.
8. Energy Consumption: It is the process to help in evades the over­
heating condition by balancing the workload transversely every nodes.

10.4 RELATED WORKS

Load balancing is a process which is utilized for distributing processing loads

to slighter dealing out nodes for developing the entire presentation of a scheme
Load-Balancing Algorithm with Overall Tasks 123

(Begum and Prashanth, 2013). In a distributed system situation, it is the devel­

opment of distributing load among a range of other nodes of an arrangement to
acquire superior entity operation and task response time (Alakeel, 2010). An
ultimate load balancing technique should keep away from overfilling or under
loading of any definite node (Moharana, Ramesh, and Powar, 2013). But, in
the case of a cloud atmosphere, the assortment of a load balancing is not purely
method due to its engagement of additional restriction like throughput, security,
dependability, etc. (Mangla and Goyal, 2017). Hence, goal of an algorithm in
cloud computing is to obtain better the response time of the task by distributing
a largely load of the method. The algorithm must also make certain that it is not
overloading any explicit node (Haryani and Jagli, 2014).
Load balancers can work in two ways: one is cooperative and non­
cooperative. In a cooperative way, all nodes work concurrently to reach
the familiar goal of optimizing the on the whole response time. In the non­
cooperative way, jobs run separately to recover the response time of local
jobs (Moharana, Ramesh, and Powar, 2013).
Cloud computing supplies a collection of duties to the customer, such
as multi-media division, online office software, online storage, etc. (Shilpa,
Kulkarni, and Kulkarni, 2014). The opportunistic load balancing algorithm
(OLB) proposes to remain distinct node busy despite the existing workloads of
every node (Saranya and Maheswari, 2015). OLB algorithm allotted schedules
to reflect nodes in arbitrary order. The minimum completion time algorithm
(MCT) disperses schedules to the nodes containing the anticipated MCT of
this schedule over other nodes (Patel, Mehta, and Bhoi, 2015). The Min-Min
scheduling algorithm (MM) presumes the identical scheduling come near as
the MCT (Kaur and Singh, 2014) algorithm to consigning a schedule, the node
to conclude this schedule with MCT over other nodes (Kaur and Patra, 2013).
The load balance min-min (LBMM) scheduling method (Rahm, 1996) imple­
ment the MM scheduling technique and load balancing stratagem and this also
evaded the unreasonably replicate project (Shidik, Azhari, and Mustofa, 2015).

10.5 PROPOSED WORK

Diversity of nodes exists in a cloud system. Each nodes cannot perform

similar tasks; therefore, just reflect on the processing of the entities is insuf­
ficient at the time of an entity is preferred to accomplish a particular assign­
ment. Thus, to pick a suitable entity to perform an assignment is awfully
significant in cloud computing.
124 Advances in Data Science and Computing Technology

The different schedule has a diverse attributes to disburse implementa­

tion. Thus, its requirements a little of the entities of definite; for case in point,
as soon as implementing individual sequence assemblage, it has to neces­
sitate toward memory enduring. To obtain the best result in the completing
of each schedule, so we will intend schedule property to imagine a dissimilar
circumstance inconsistent in which it is according to an entity of schedule
requirement to deposit resolution variable.

The process of our approach is presented as follows:

 Step 1: Respective process to calculate the total execution time of every
schedule for all entities, correspondingly.
 Step 2: It is to acquire the schedule having the greatest execution time.
 Step 3: It is to locate the entity which is not dispersed and has the least
execution time for the schedule preferred in Step 2. Then, respective
schedule is stimulated to the elected entity for computation purpose.
 Step 4: If, in step 2, no unallocated entity can be selected, then that all entity,
together with unallocated and allocated entities, should be reassessed. The
smallest amount execution time of an allocated entity is the summation of
the smallest amount completion time of the consign schedule on this entity
and the smallest amount execution time of the existing schedule. The lowest
amount execution time of an unallocated entity is the existing smallest
amount execution time for the schedule. It is to determine the unallocated
entity or allocated entity having the lowest amount execution time for the
schedule selected in Step 2 afterward, this schedule is stimulated to the
elected entity for computation.
 Step 5: Perform once more Step 2 to Step 4 up to all schedules have been
accomplished absolutely.

10.6 ILLUSTRATION

In the subsequent part, an example to be computed by means of the projected

algorithm is going behind.
Schedule Entity N11 N12 N13 N14
T1 13 14 11 15
T2 17 25 14 26
T3 27 32 13 34
T4 18 25 19 32
Load-Balancing Algorithm with Overall Tasks 125

 Step 1: It is to estimate the total execution time of distinct schedule for

all entities, individually.
Schedule Entity N11 N12 N13 N14 Total
Schedule
T1 13 14 11 15 53
T2 17 25 14 26 82
T3 27 32 13 34 106
T4 18 25 19 32 94

 Step 2: Then locate the schedule that has the greatest execution time.
Schedule Entity N11 N12 N13 N14 Total
Schedule
T1 13 14 11 15 53
T2 17 25 14 26 82
T3 27 32 13 34 106
T4 18 25 19 32 94

 Step 3: Next to locate the unallocated entity which has the least execu­
tion time for the schedule preferred in Step 2. Then, this schedule is
stimulated to the elected entity for computation.
Schedule Entity N11 N12 N13 N14 Total
Schedule
T1
13 14 11 15 53
T2 17 25 14 26 82

T3 27 32
13 34 106
T4 18 25 19 32 94

 Step 5: Perform over again Step 2 to Step 4 up to all schedules have been
executed totally.
Schedule Entity C11 C12 C13 C14 Total
Schedule
T1 13 14 11 15 53
T2 17 25 14 26 82
T3 27 32 13 34 106
T4 18 25 19 32 94
126 Advances in Data Science and Computing Technology

Final Result
Schedule Entity C11 C12 C13 C14
T1 13 14 11 15
T2 17 25 14 26
T3 27 32 13 34
T4 18 25 19 32

10.7 COMPARISON

Figure 10.1 illustrates the execution time for distinct schedule at several
computing entities. To analyze the performance of our projected method is
evaluated with a new method by the case shown in Figure 10.1. Figure 10.1
demonstrates the assessment of the execution time of distinct computing
entity among our method. Respective completion times for ultimate all
schedules by means of the proposed technique, LBMM, and MM, are 24, 33,
and 35 ms, respectively. Our technique attains the MCT and enhanced load
balancing than AK algorithms.

FIGURE 10.1 The assessment of the completion time of distinct schedule at the several
entities.
Load-Balancing Algorithm with Overall Tasks 127

10.8 CONCLUSION

In our chapter, we projected a competent scheduling technique, LBOT,

for the cloud computing environment to dispense schedule to work out
entities according to their respective resource competence correspondingly,
our method can attain superior load balancing and performance than other
existing method, such as LB3M, LBMM, and MM, from the case history.

KEYWORDS

• cloud computing
• distributed system
• load balancing

REFERENCES

Adhikari, J., & Patil, S., (2012). Load balancing the essential factor in cloud computing.
International Journal of Engineering Research & Technology (IJERT).
Akyildiz, I. F., Janise, M., Loren, C. M., Ramon, P., & Yelena, Y., (1999). Medium access
control protocols for multimedia traffic in wireless networks. IEEE Network, 13(4), 39–47.
Alakeel, A. M., (2010). A fuzzy dynamic load balancing algorithm for homogenous distributed
systems. International Journal of Computer Science and Network Security, 10(6), 153–160.
Alnusairi, T. S., Shahin, A. A., & Daadaa, Y., (2018). Binary PSOGSA for Load Balancing
Task Scheduling in Cloud Environment. arXiv preprint arXiv:1806.00329.
Ansari, M. B., & Sachin, P., (2015). Load balancing and issue monitoring of cloud nodes.
International Journal, 5(7).
Begum, S., & Prashanth, C. S. R., (2013). Review of load balancing in cloud computing.
International Journal of Computer Science Issues (IJCSI), 10(1), 343.
Bhargava, S., & Goyal, S., (2013). Dynamic load balancing in cloud using live migration of
virtual machines. Computing, 3(7), 8.
Bittencourt, L. F., Miyazawa, F. K., & Vignatti, A. L., (2012). Distributed load balancing
algorithms for heterogeneous players in asynchronous networks. Journal of Universal
Computer Science.
Chan, M. C., Jiang, J. R., & Huang, S. T., (2012). Fault-tolerant and secure networked storage.
In: Digital Information Management (ICDIM), 2012 Seventh International Conference (pp.
186–191). IEEE.
Chen, C. H., Lin, H. F., Chang, H. C., Ho, P. H., & Lo, C. C., (2013). An analytical framework
of a deployment strategy for cloud computing services: A case study of academic websites.
Mathematical Problems in Engineering, 2013.
128 Advances in Data Science and Computing Technology

Dinh, H. T., Lee, C., Niyato, D., & Wang, P., (2013). A survey of mobile cloud computing:
Architecture, applications, and approaches. Wireless Communications and Mobile Computing,
13(18), 1587–1611.
Eager, D. L., Lazowska, E. D., & Zahorjan, J., (1985). A comparison of receiver-initiated
and sender-initiated adaptive load sharing. ACM SIGMETRICS Performance Evaluation
Review, 13(2), 1–3.
Gupta, E., & Deshpande, V., (2014). A technique based on ant colony optimization for load
balancing in cloud data center. In: Information Technology (ICIT), 2014 International
Conference on (pp. 12–17). IEEE.
Haryani, N., & Jagli, D., (2014). A dynamic method for load balancing in cloud computing.
IOSR Journal of Computer Engineering (IOSR-JCE), 16(4), 23–28.
Jain, R., & Paul, S., (2013). Network Virtualization and Software Defined Networking for
Cloud Computing: A Survey (Vol. 51, No. 11, pp. 24–31). IEEE Communications Magazine.
Jeevitha, M., Chandrasekar, A., & Karthik, S., (2015). Survey on verification of storage
correctness in cloud computing. International Journal of Engineering and Computer
Science, 4(09).
Kabakus, A. T., & Kara, R., (2017). A Performance evaluation of drop box in the light of
personal cloud storage systems. International Journal of Computer Applications, 163(5).
Kaur, D., & Singh, S., (2014). An efficient job scheduling algorithm using min-min and ant
colony concept for grid computing. International Journal of Engineering and Computer
Science, 3(07).
Kaur, R., & Patra, P. K., (2013). Resource allocation with improved min-min algorithm.
International Journal of Computer Applications, 76(15).
Krakowiak, S., (2009). Distributed under a Creative Commons license. Middleware Architecture
with Patterns and Frameworks. http://creativecommons.org/licenses/by-nc-nd/3.0/ (accessed
on 10 January 2022).
Kumar, K., & Lu, Y. H., (2010). Cloud computing for mobile users: Can offloading computation
save energy?. Computer, 43(4), 51–56.
Kumar, R., Agrawal, G., Theobald, K., Zoppetti, G. M., & Gao, G. R., (2001). Compiling
several classes of reductions on a multithreaded architecture. In: Proceedings of Mid-Atlantic
Student Workshop on Programming Languages and Systems.
Mangla, P., & Goyal, S. K., (2017). Study of various heuristic approaches in cloud computing.
International Journal of Advanced Research in Computer Science, 8(3).
Mani, V., Suresh, S., & Kim, H. J., (2005). Real-coded genetic algorithms for optimal
static load balancing in distributed computing systems with communication delays. In:
International Conference on Computational Science and its Applications (pp. 269–279).
Springer, Berlin, Heidelberg.
Mansouri, Y., & Buyya, R., (2016). To move or not to move: Cost optimization in a dual cloud-
based storage architecture. Journal of Network and Computer Applications, 75, 223–235.
Mishra, N. K., & Mishra, N., (2015). Load balancing techniques: Need, objectives, and major
challenges in cloud computing-a systematic review. International Journal of Computer
Applications, 131(18).
Moharana, S. S., Ramesh, R. D., & Powar, D., (2013). Analysis of load balancers in cloud
computing. International Journal of Computer Science and Engineering, 2(2), 101–108.
Moharana, S. S., Ramesh, R. D., & Powar, D., (2013). Analysis of load balancers in cloud
computing. International Journal of Computer Science and Engineering, 2(2), 101–108.
Load-Balancing Algorithm with Overall Tasks 129

Moura, J., & Hutchison, D., (2016). Review and analysis of networking challenges in cloud
computing. Journal of Network and Computer Applications, 60, 113–129.
Padhy, R. P., (2012). Virtualization techniques & technologies: State-of-the-art. International
Journal of Global Research in Computer Science (UGC Approved Journal), 2(12), 29–43.
Pareek, P., (2013). Cloud computing security from single to multi clouds using secret sharing
algorithm. International Journal of Advanced Research in Computer Engineering &
Technology, 2(12), 3261–3264.
Patel, G., Mehta, R., & Bhoi, U., (2015). Enhanced load balanced min-min algorithm for
static meta task scheduling in cloud computing. Procedia Computer Science, 57, 545–553.
Patil, U., & Shedge, R., (2013). Improved hybrid dynamic load balancing algorithm for
distributed environment. International Journal of Scientific and Research Publications,
3(3), 1.
Pearce, M., Zeadally, S., & Hunt, R., (2013). Virtualization: Issues, security, threats, and
solutions. ACM Computing Surveys (CSUR), 45(2), 17.
Rahm, E., (1996). Dynamic load balancing in parallel database systems. In: European
Conference on Parallel Processing (pp. 37–52), Springer, Berlin, Heidelberg.
Rastogi, D., Bansal, A., & Hasteer, N., (2013). Techniques of load balancing in cloud
computing: A survey, In: International Conference on Computer Science and Engineering
(CSE) (pp. 140–143).
Sabahi, F., (2011). Security of virtualization level in cloud computing. In: Proc. 4th Intl. Conf.
on Computer Science and Information Technology (pp. 197–201). Chengdu, China.
Saranya, D., & Maheswari, L. S., (2015). Load balancing algorithms in cloud computing: A
review. International Journal of Advanced Research in Computer Science and Software
Engineering Research Paper.
Shidik, G. F., Azhari, M. K., & Mustofa, K., (2015). Evaluation of selection policy with
various virtual machine instances in dynamic VM consolidation for energy efficient at
cloud data centers. JNW, 10(7), 397–406.
Shilpa, S., Kulkarni, S., & Kulkarni, S., (2014). Analysis of load balancing algorithms in
cloud computing and study of game theory [J]. Analysis, 3(4), 1818–1823.
Sibiya, G., Venter, H. S., & Fogwill, T., (2015). Digital forensics in the cloud: The state of the
art. In: IST-Africa Conference, 2015 (pp. 1–9). IEEE.
Sran, N., & Navdeep, K., (2013). Comparative Analysis of Existing Load balancing techniques
in cloud computing. International Journal of Engineering Science Invention, 2(1), 60–63.
Vakkalanka, S., (2012). Classification of job scheduling algorithms for balancing load on web
servers. International Journal of Modern Engineering Research (IJMER), 2(5), 3679–3683.
Wang, L., Gregor, V. L., Andre, Y., Xi, H., Marcel, K., Jie, T., & Cheng, F., (2010). Cloud
computing: A perspective study. New Generation Computing, 28(2), 137–146.
Willebeek-LeMair, M. H., & Reeves, A. P., (1993). Strategies for dynamic load balancing on
highly parallel computers. IEEE Transactions on Parallel and Distributed Systems, 4(9),
979–993.
Youseff, L., Butrico, M., & Da Silva, D., (2008). Toward a unified ontology of cloud
computing. In: Grid Computing Environments Workshop, 2008, GCE’08 (pp. 1–10). IEEE.
Zhang, F., Chen, J., Chen, H., & Zang, B., (2011). CloudVisor: Retrofitting protection of
virtual machines in multi-tenant cloud with nested virtualization. In: Proceedings of the
Twenty-Third ACM Symposium on Operating Systems Principles (pp. 203–216). ACM.
CHAPTER 11

IMPLEMENTATION OF LOAD
BALANCING OF MATRIX PROBLEM
WITH THE ROW SUM METHOD
PAYEL RAY, ENAKSHMI NANDI, RANJAN KUMAR MONDAL, and
DEBABRATA SARDDAR
Computer Science and Engineering, University of Kalyani,
West Bengal, India, E-mail: [email protected] (P. Ray)

ABSTRACT

Cloud computing is turning into one of the most emerging technologies

based on the internet. In the simplest terms, cloud computing means
storing and gaining access to data and application over the internet as an
alternative to your computer’s hard drive. Cloud computing is the delivery
of on-demand computing services from applications to storage and deals
with power typically over the internet and on a pay-as-you-go basis. There
is a wide range of network servers joined to online computing to provide
different types of online services to cloud clients. There are some fewer
figures of network servers having joined the cloud networks that have to
perform greater tasks at a limited time. So, it is certainly easy to accomplish
the entire works at a specific time. Few systems complete each task, so there
is a demand for balancing workloads at a limited time. Here Load balancing
reduces the time limit of each task in an exacting process.
There are invariably no chances to continue an equal quantity of network
servers to accomplish the same quantity of jobs in the same period. All tasks
to be performed would be higher than the connection to the servers. There
are inadequate servers that have to hand over a million jobs in an instant.
132 Advances in Data Science and Computing Technology

We will propose an algorithm that insufficient nodes perform the jobs

here; some jobs are much more than the nodes and balance each node to
improve the quality of services in cloud computing.

11.1 INTRODUCTION

Cloud computing is the exploitation of online sources accepted as a service over

an arrangement (Chan et al., 2012). The name emerges from the usual make
use of a cloud-shaped sign as an abstraction for the system communications
containing in system figures (Jeevitha, Chandrasekar, and Karthik, 2015).
The Cloud system entrusts distant services with a client’s data, program, and
computation. Cloud computing comprises resources made last on the web as
provided by third-party services. These services serve the perfection of the
entrance to current software functions and high-end networks of physical
workstation machines (Sarddar and Ray, 2017). The purpose of cloud
computing is to treat traditional device or high-performance computing
practice, again, and repeatedly made use by research resources, to execute
trillions of calculations per unit time, in consumer-oriented applications for
example financial portfolios, to deliver personalized guidance, to allow data
storage or to power extensive, immersive physical machine games. The cloud
computing uses methods of comprehensive collections of workstations,
running low-cost customer PC technology with unique connections to extend
data-handling chores across them. This distributed infrastructure encloses
enormous pools of systems connected. It requires virtualization techniques
to handle the effect of cloud computing.

11.2 CLOUD COMPUTING AND LOAD BALANCING

11.2.1 CLOUD SYSTEM

Cloud system (cloud computing) is a method for enabling expedient,

on-demand network right off the entrance to a shared pool of configurable
resources competent to be quickly stipulation and issued with the slightest
management attempt (Moura and Hutchison, 2016). It bases the cloud system
on virtualization. Virtualization is the major feature of the cloud system. The
cloud system virtualizes an individual system into several virtual systems
(Ray and Sarddar, 2018). A virtual machine is a software implementation of
the physical machine. A low-level program called virtual machine monitor
Implementation of Load Balancing of Matrix Problem 133

is reliable for the sharing of an individual physical machine among different

resources.
Cloud computing aims to enhance the function of the system and to
perform the most essential resource expenditure, greatest throughout, least
response time and avoiding extra load.

11.3 CLASSIFICATION OF ALGORITHMS

It has divided the load balancing algorithms based on the recent state of a
system and who started the process:
1. Depending on which client initiates the process (Eager, Lazowska,
and Zahorjan, 1985):
i. Sender-Initiated: Sender or client starts the completion of the
load balancing algorithm on analyzing the need for load balancing.
ii. Receiver-Initiated: Receiver or workstation starts the finaliza­
tion of a load balancing algorithm on diagnosing the demand for
load balancing.
iii. Symmetric: This category of the algorithm is a merging of
the sender-initiated category and receiver-initiated category
algorithms.
2. Depending on the present state of the static system algorithm: In
the static algorithm, there is a uniform distribution of traffic among
the workstations. This algorithm requires a prior recognition of the
resources of the system so that the judgment of moving the loads not
depend on the presented situation of the system.
3. Dynamic Algorithm: In the dynamic algorithm, for balancing the
load, the lightest workstation in the entire process is considered and
selected. For this, real-time connection with the Internet is required
that can expand the data transfer in the system. Here to determine to
manage the load, it capitalizes on the living shape of the system.

11.4 LOAD BALANCING IN CLOUD

Load balancing is a process of reassigning the entire load to the dissimilar

nodes of the distributed system to form resource use proper-formed and to
gain an enhanced response time of the job, alongside removing circumstances
where a few of the nodes are extra loaded whereas several other nodes are
134 Advances in Data Science and Computing Technology

loaded easily. A load-balancing method strives to build up the application

of resources with a light load or idle resources efficiently by freeing the
resources with an extra load. The process seeks to handle the load amongst
each obtainable resource. It strives to reduce the make-span of the successful
exploitation of sources.
In distributed processes, comprised of homogeneous and distributed
resources, it has researched load balancing processes extensively. These
processes will not be living in cloud architecture because of its heterogeneity,
scalability, and freedom. This creates load-balanced scheduling for cloud
computing extra high and a provocative issue for many analysts.
The Non-traditional methods differ from the conventional process because
it brings about the most beneficial results in a concise hour. There is no most
excellent scheduling process for a cloud system. The alternative opportunity
is to determine a convenient approach to apply in known cloud surroundings
owing to the factors of the tasks, servers, and network heterogeneity.

11.5 RELATED WORKS

1. LBMM (Kokilavani and Amalarethinam, 2011): It is a static

balancing algorithm. This algorithm uses load balancing amongst
machines considering it. The Min-Min algorithm demonstrates the
lowest completion time for all jobs. Then it upholds the tasks with the
least completion time amongst all the tasks. This algorithm follows by
permitting the positions to the process design the lowest makespan.
Min – it schedules Min states similar procedures in the prospect
of all jobs. The algorithm demonstrates the Min-Min schedule and
creates an opinion of the node with the drawing makespan. Conse­
quent to the system requires the task with the least execution time. It
is referring to the execution time of the chosen job for each source.
It considers the peak execution time of the elected a job with the
make-span. If the execution time is smallest, therefore, it shares the
preferred job to the machine, holding the maximum makespan. Else,
the latest maximum makespan of the work is preferred, and the steps
are consistent. The program tests if it assigns each system with each
task. In these conditions where the figure of undersized jobs is higher
than the number of abundant jobs in meta-job, these algorithms have
improved appearance. This algorithm does not recognize light and
upgrade systems, heterogeneity, and jobs.
Implementation of Load Balancing of Matrix Problem 135

2. Two-Phase Scheduling (Mishra, Sahoo, and Parida, 2018): It is the

formation of OLB and LBMM algorithms to make superior comple­
tion competency and moves on with the process load balancing. OLB
scheduling keeps on each process in managing situations to have a
hold of the purpose of load balancing with the LBMM algorithm
is maintained to diminish the completion of the time of all jobs on
the system by this means diminishing the complete execution time.
This algorithm was working on having rescued from the process of
sources and bought the effort competency.
3. Min-Min (Shaw and Singh, 2014): It assumes MM with a load
balancing strategy. In this approach, it splits all tasks into subtasks.
The completion times of each subtask on every server, and the
threshold, are measured. All subtasks, along with the server to raises
the shortest volume of execution time for the subtask goes into Min
time. The subtask having the least amount of execution time amongst
all subtasks is selected and gave away to the associated server.
4. LB3M (Madni et al., 2017): This scheme determines the average
completion time of all subtasks on all servers. The subtask has the
largest average completion time, and the server is making the least
completion time determined. At present, the assigned server will deal
with the chosen subtask. If the server is before assigned, formerly
the procedure evaluates the completion time of the server. For the
designated server, make-span is the summation of the makespan of
appointed tasks and completion time for the current subtask. For the
unallocated server, make-span is the execution time of the remaining
task. This proceeding is continuous in the subsequently unassigned
subtasks.

11.6 PROPOSED METHOD

To determine the matrix cost in addition to a combination of the job(s)

vs. machine(s) of an unbalanced matrix, to concentrate on a problem
consisting of machines M={M1 , M2, …, Mm}. The jobs J = {J1, J2, …, Jn},
is considered to be assigned for execution on the ‘m’ obtainable machines
with the implementation cost Cij, where i = 1, 2, …, m, and j = 1, 2, …, n
are mentioned in the matrix in which m is greater than n. First of all, we
get the sum of all rows and all columns of the matrix, store the results in
the array, row-sum, and column-sum. Then to choose the first n rows by
136 Advances in Data Science and Computing Technology

row-sum, i.e., starting with smallest to next smallest to the array row-sum
and deleting rows related to the remaining (r) jobs. Store the results in
the new array that should be the array for the first sub-problem. Do this
process again until remaining jobs become less than a machine when
remaining jobs are less than n, then, deleting (c) columns by column-sum,
i.e., corresponding to the value(s) most higher to next higher to form the
last sub-problem. Accumulate the results in the new array that will be the
array for the last sub-problem. Apply the Hungarian method (Dive, Ingale,
and Shahade, 2013) to get the best possible solution of all sub-problems,
which is now becoming a balanced problem. Lastly, rearrange all the sub­
problems to acquire the best possible cost.

11.6.1 ALGORITHM

To present an algorithmic illustration of the method, consider a problem that

consists of ‘m’ machines M = {M1, M2, …, Mm}. Moreover, ‘n’ jobs J =
{J1, J2, …, Jn} is considered to be assigned for execution on ‘m’ obtainable
machines with the implementation cost Cij, where i = 1, 2, …, m, and j = 1,
2, …, n, where m>n, i.e., the jobs is greater than machines.

 Step 1: Input: m×n matrix.

 Step 2: To calculate the row sum of all tasks, correspondingly.
 Step 3: To divide the problem into two components. The first component
will be the n×n matrix based on the highest row-sum, and the remaining
components would be the n×m format where n>m. To calculate easy, we
follow this rule (First, the component will be followed by a single machine
with a single job with executes shortest execution time based on highest
row-sum, and the second component will be followed by a device that
performs minimum execution time based on maximum row-sum.).
 Step 4: In the second component, find least cost of unassigned jobs Ji of all
already assigned machines from highest row-sum, and it is to be attached to
the related machine and remove from the assigned job and related machine
from the matrix and find next least cost of unassigned jobs from next
maximum row-sum and assign the task to its matching machine and so on in
anticipation of each task assigned to their consequent device.
 Step 5: So, each job assigned as a minimum to its related machine and more
than two jobs are not allocated to any device.
 Step 8: Stop.
Implementation of Load Balancing of Matrix Problem 137

11.7 ILLUSTRATION OF AN EXAMPLE

Let us a set that has 5 machines M = {M1, M2, M3, M4, M5}, and 8 jobs J =
{J1 , J2, J3, J4, J5, J6, J7, J8}. The set contains the execution costs of every job
to each machine.

 Step 1: Input: 5×8 matrix.

Jn/Mm M1 M2 M3 M4 M5
J1 151 277 185 276 321
J2 245 286 256 264 402
J3 246 245 412 423 257
J4 269 175 145 125 156
J5 421 178 185 425 235
J6 257 257 125 325 362
J7 159 268 412 256 286
J8 365 286 236 314 279

 Step 2: It is to calculate the Row Sum of each task, respectively.

Jn/Mm M1 M2 M3 M4 M5 Row-Sum
J1 151 277 185 276 321 1,210
J2 245 286 256 264 402 1,453
J3 246 245 412 423 257 1,583
J4 269 175 145 125 156 870
J5 421 178 185 425 235 1,444
J6 257 257 125 325 362 1,326
J7 159 268 412 256 286 1,381
J8 365 286 236 314 279 1,480

 Step 3: We split the matrix into two parts. The first part will be the n×n
matrix based on the maximum row-sum, and the remaining parts would
be the n×m format. Easy to calculate, we follow this rule. First, the part
will be followed by one machine with one job with executes minimum
execution time based on maximum row-sum, and the second part will be
followed by a device that performs minimum execution time based on
maximum row-sum.
138 Advances in Data Science and Computing Technology

Jn/Mm M1 M2 M3 M4 M5 Row-Sum
J2 245 286 256 264 402 1,453
J3 246 245 412 423 257 1,583
J5 421 178 185 425 235 1,444
J7 159 268 412 256 286 1,381
J8 365 286 236 314 279 1,480

 Step 4: Like all machines assigned by jobs by now, thus we want to

remain the lowest cost of unassigned jobs in the employed system. Find
the minimum value of unassigned jobs Ji of all already assigned devices
from maximum row-sum, and it is to be attached to the corresponding
machine and remove from the assigned job and similar machine from the
list and find next minimum cost of unassigned jobs from next maximum
row-sum and allocate the task to its corresponding machine and so on
until all tasks assigned to their corresponding device.

Jn/Mm M1 M2 M3 M4 M5 Row-Sum
J1 151 277 185 276 321 1,210
J4 269 175 145 125 156 870
J6 257 257 125 3255 362 1,326

 Step 5: So, all jobs assigned at least to its corresponding machine, and
more the two jobs are not allocated to any device.
Jn/Mm M1 M2 M3 M4 M5
J1 151 277 185 276 321
J2 245 286 256 264 402
J3 246 245 412 423 257
J4 269 175 145 125 156
J5 421 178 185 425 235
J6 257 257 125 325 362
J7 159 268 412 256 286
J8 365 286 236 314 279

Final Result
M1 → J1*J2 → 151 + 245 = 396
M2 → J3 → 245
Implementation of Load Balancing of Matrix Problem 139

M3 → J6*J8 → 125 + 236 = 361

M4 → J4*J7 → 125 + 256 = 381

M5 → J5 → 235

Our approach: 1,618

Mm M1 M2 M3 M4 M5
Costs 396 245 361 381 235

11.8 RESULT ANALYSIS

The experiments were performed by following table execution time (ms)

of each task at all nodes. The version of the system is Intel Core i3, 4th
Generation processor, 3.4 GHz CPU and 4 GB RAM running on Windows 7
platform. We have considered makespan and we have conducted two sets of
simulation scenarios as follows.
Figure 11.1 explains the implementation time for each task at several
computing nodes. To compute the presentation of our proposed work is
evaluated with other processes shown in Figure 11.1. Figure 11.1 displays
the evaluation of the execution time of all nodes among in our approach.
For completing each task by using the proposed algorithm, LBMM, HM,
and MM, the makespan is 396, 423, 461, and 555 ms, correspondingly. Our
approach attained the lowest completion time and improved balancing of the
load than other algorithms.

11.9 CONCLUSION

It assigns None of the works to dummy machines. The result of the unbal­
anced matrix problem, the matrix obtained with the back of the Hungarian
method, is pointed out. It appoints all subtasks with makespan to its related
system so that the result is farthest desirable for the system. Although some
devices perform over one job, makespan is most convenient, and loads are
adjusting with our proposed work.
The technique is presented in an algorithmic shape and implemented on
the many sets of input data to test the act and usefulness of the algorithm.
Applying to load balancing is straightforward.
140 Advances in Data Science and Computing Technology

FIGURE 11.1 Execution time (ms) of each task at different computing nodes.

KEYWORDS

• cloud computing
• load balancing
• minimum completion time

REFERENCES

Chan, M. C., Jiang, J. R., & Huang, S. T., (2012). Fault-tolerant and secure networked storage.
In: 2012 Seventh International Conference on Digital Information Management (ICDIM)
(pp. 186–191). IEEE.
Dive, S. R., Ingale,A. D., & Shahade, M. R., (2013). Virtual machine and network-performance,
study, advantages, and virtualization option. International Journal of Advanced Research in
Computer Science, 4(6).
Eager, D. L., Lazowska, E. D., & Zahorjan, J., (1985). A comparison of receiver-initiated
and sender-initiated adaptive load sharing. ACM SIGMETRICS Performance Evaluation
Review, 13(2), 1–3.
Jeevitha, M., Chandrasekar, A., & Karthik, S., (2015). Survey on verification of storage
correctness in cloud computing. International Journal of Engineering and Computer Science,
4(9).
Implementation of Load Balancing of Matrix Problem 141

Kokilavani, T., & Amalarethinam, D. G., (2011). Load balanced min-min algorithm for static
meta-task scheduling in grid computing. International Journal of Computer Applications,
20(2), 43–49.
Madni, S. H. H., Latiff, M. S. A., Abdullahi, M., & Usman, M. J., (2017). Performance
comparison of heuristic algorithms for task scheduling in the IaaS cloud computing
environment. PloS One, 12(5), e0176321.
Mishra, S. K., Sahoo, B., & Parida, P. P., (2018). Load balancing in cloud computing: A big
picture. Journal of King Saud University-Computer and Information Sciences.
Moura, J., & Hutchison, D., (2016). Review and analysis of networking challenges in cloud
computing. Journal of Network and Computer Applications, 60, 113–129.
Ray, P., & Sarddar, D., (2018). Load balancing with inadequate machines in cloud computing
networks. In: Mandal, J., & Sinha, D., (eds.), Social Transformation – Digital Way (Vol.
836). CSI 2018. Communications in computer and information science. Springer, Singapore.
Sarddar, D., & Ray, P., (2017). A new effort for distributed load balancing. International
Journal of Grid and Distributed Computing, 10(8), 1–9.
Shaw, S. B., & Singh, A. K., (2014). A survey on scheduling and load balancing techniques in
a cloud computing environment. In: Computer and Communication Technology (ICCCT),
2014 International Conference (pp. 87–95). IEEE.
 PART IV

Advances in Embedded System-Based

Applications

CHAPTER 12

MICROCONTROLLER-BASED
WEARABLE LOCATION TRACKER
SYSTEM WITH IMMEDIATE SUPPORT
FACILITY
SUMANTA CHATTERJEE, DWAIPAYAN SAHA,
INDRANI MUKHERJEE, and JESMIN ROY
Department of Computer Science and Engineering,

JIS College of Engineering, Kalyani, West Bengal, India,

E-mail: [email protected] (S. Chatterjee)

ABSTRACT

The work commitments are commendable but along with this security also
comes together and as much dedications are there; their lives are also at risk
and so it is our responsibilities to take care of them. So, a location tracker
device is proposed which is named as “wearable location tracker” and it
is basically a device in any form such as pins, wearable watches or even
customized shirts which gives the real time location of the person and there
are some special features also added to it such as the pulse rate count and
the body temperature (BT) of the people. This chapter is mainly based on
how one can save people from danger and provide them immediate help to
evacuate them from danger. This device not only gives location but it has
the facility of SOS system for sending message to near one’s and emergency
contact to provide immediate help.
146 Advances in Data Science and Computing Technology

12.1 INTRODUCTION

Security is one of the most important and highly concerned aspects of every
one’s life. As the overall circumstances are not in a good scenario along with
security, safety for women and children are also gaining concern as the days
are passing. Along with safety, tracking is also as major concern as more
number of crimes on children and women are reported nowadays.
So, every near and dear one is of them want to monitor their activities
and want to keep track of their loved ones and over the phone is not always
possible. Taking precautions and preventions is always better than meeting
with an accident. So, our proposed idea will serve the best for this purpose
which is built using microcontroller. A system in a form of a chip that does
a job can simply be called a microcontroller. It contains a small amount of
RAM, an integrated processor, which are very useful while interacting with
things connected to the microprocessor. And so, as our product too have.
Various parameters are used along with the microcontroller for making this
wearable tracker and advanced technologies are used to their uttermost to
serve the best to the mankind.
Our proposed idea mainly focuses on making a wearable tracker device
when people will be facing any danger, they can take help by using our
proposed device just by tapping the SOS button to request for immediate
help and in case of people working in any fields the device will monitor their
pulse rate and their vital body parameters and accordingly if anything goes
wrong the device will send signals for immediate arrangements for rescue.
This proposed device can be of great help for the safety measurements of
women and kids as well. This device will be of great help for people to
provide them immediate help. Thus, the device is of great importance for
every aspect people.

12.2 RELATED WORK

There are various existing plans regarding location tracker device. Some of
them are as in subsections.

12.2.1 SMART IOT DEVICE FOR CHILD SAFETY AND TRACKING

This chapter deals with proposal of system or device which will track
children’s activity and will update constantly with the help of internet
Microcontroller-Based Wearable Location Tracker System 147

of things (IoT). IoT devices are smart and advanced devices, which are
able to take decisions by sensing the environment surrounding the device.
IoT connects billions of devices together with the help of the internet
connections and it has the ability to transfer large number of data over
a network without requiring human-to-human or human-to-computer
interaction. IoT is an emerging field now and it is gaining a lot of
popularity also.
The proposed system has the facility of camera, GPS system, hear beat
sensor and touch sensor which can monitor and track their children’s activity
and they can keep a record of them (Figure 12.1).

Rechargeable
Battery

Temperature
sensor

Touch sensor LinkIt ONE

Board Cloud

Serial camera

Heart Beat
Sensor Mobile
1

Mobile Mobile
2 3

GSM GPS
Antenna Antenna

MMS
and SMS
to Mobile

FIGURE 12.1 The workflow diagram of the IoT device for child safety and tracking.

12.2.2 GPS AND SMS-BASED CHILD TRACKING SYSTEM USING

SMART PHONE

This chapter mainly focusses on safety of child. The authors have here
proposed a solution based on Android which will help them to track the real
time location of their wards. Recently, the smart phones are well equipped
148 Advances in Data Science and Computing Technology

with sensors that are capable in doing this task of giving the location of the
user.
Since the services are provided by smart phones, so the solution presented,
takes the advantage of that to provide the required services. Hence real time
location is obtained using the sensors like GPS and GSM. The system has to
parts parents’ part and children part. The parent’s side will send a request to
the child’s device to send the location of the child, and the child device will
acknowledge the same and send a reply.

12.2.3 SMART WEARABLE BLUETOOTH FITNESS TRACKER

In this proposed chapter, they have decided to incorporate electronic

technology into the shoes that can provide data with respect to the fitness
record thus measuring the calories. This device is used by sportspersons,
athletes, or normal people who are fitness freak. This device is mainly
planned to have a record of fitness of the people.
Their proposed device uses an accelerometer-based chip that records and
gives accurate data about the movement and the same sensor also records the
body temperature (BT) and accordingly necessary steps will be taken.

12.2.4 IoT BASED WEARABLE SMART HEALTH MONITORING

SYSTEM

In this chapter, device is proposed for individuals with heart disease. Heart
rate variability (HRV), heart rate (HR), and BT values are considered as
a vital signs that must be measured regularly in case heart patients. The
proposed solution is an android-based application, which is developed to
monitor HR, HRV, and CT parameters for patients.
The proposed measuring system consists of wearable sensors which
constantly measures patient signs. Then immediately send the measured
signals to android interface via wireless connection. If the earlier recorded
critical values for the patient have exceeded, the HR, HRV, CT values and
along the real time location of patient is sent both to family members and
doctor as e-mail and twitter notification. This particular wearable system
helps the users to move around without any fear. It also ensures that e very
citizen feels safe wherever they go.
Microcontroller-Based Wearable Location Tracker System 149

12.2.5 TEMPERATURE REGULATING TIMEPIECES

This chapter worked with body parameters to track people in need. This
device uses BT sensor to record the vital body parameters and provide
immediate help to the victim.
And lots of other tracking devices are available which used GPS and
GSM module to track the person who needs help in their workplace or in
personal area of interest. And with the help of GSM module, they can send
help message to the contacts registered to the GSM module and easily imme­
diate help can be given to them.

12.2.6 PEDESTRIAN TRACKING WITH SHOE-MOUNTED

INERTIAL SENSORS

In this chapter, authors mainly focused on delivering a system to rescue fire­

fighters or other emergency responders. They developed a new system named
as NavShoe, which uses a new and advanced approach for position tracking
based on inertial sensing. The wireless inertial sensor is very small enough
to easily tuck into the shoelaces, and sufficiently have low power to run all
day on a small battery. It cannot be used alone with precise registration for
close-range objects, but in applications in outdoor augmenting distant objects
can be used and a user would rarely notice the NavShoe’s meter-level error
which is combined with any of the error in the head’s assumed location is
relative to the foot. NavShoe can hugely reduce the database search space for
computer vision, which makes it much simpler and more robust. The NavShoe
device not only provides robust approximate position, but it also provides an
extremely accurate orientation tracker on the foot (Figures 12.2 and 12.3).
There are various tracking devices which are used for tracking logistic
and tracking various work field people. Thus, they are of great importance
for helping mankind.

12.3 PROPOSED IDEA

Leena, living in a metropolitan city, is walking an along a lonely street, it is

dark, late at night, the street has dim light. Her parents are tensed, waiting,
and praying for her to come home safely. Her father is bed ridden, her brother
studies in a school, and she is working to earn for her family. She reaches
150 Advances in Data Science and Computing Technology

her home safely, at 10.30 p.m. Despite living in a metropolitan city, her
parents are praying so that their daughter reaches home safely. This concern
pinpoints and tells us that her daughter is not safe in her own city.

FIGURE 12.2 Proposed diagram of the pedestrian tracking with shoe-mounted inertial sensors.

FIGURE 12.3 A proposed idea in the form of a demo product.

Microcontroller-Based Wearable Location Tracker System 151

Shivu comes home alone from his school. As long as he does not come
back, there are a lot of tension and negative thoughts regarding his safety
that come across the mind of his mom. An innocent child is not safe either.
In today’s world, security, and safety are the biggest concerns for an indi­
vidual, especially for women and children. They are kidnapped, trafficked,
physically abused, harassed, and what not. There are a lot of instances when
our near and dear ones want keep a track of our location so that they can
monitor us and feel assured about our safety. But is it only enough to get a
location of an individual? How will someone get to know whether an indi­
vidual is in a danger or need any help?
For this, a product has been devised which is the solution to all these
problems. The product is a wearable device which can be in any form of
a wearable device, pins, or even custom shirts. The system will not only
monitor the real time location but also will have the facilities to send
alert at the time of emergency and need. This system will also help in
mitigating tracking problem even if a fisherman loses his boat in the sea
(Figure 12.4).

POWER SUPPLY

GSM MODULE
ARDUINO SERVER
MICRO-
GPS MODULE CONTROLLER

FIGURE 12.4 Block diagram of the proposed system.

12.4 HARDWARE REQUIREMENT

In order to make the proposed idea feasible and make such a device which
has been proposed above, a few necessary components are needed which will
give life to the proposed device in reality. Below a list of required hardware
has been given along with their specification and utility:
152 Advances in Data Science and Computing Technology

12.4.1 ARDUINO UNO MICROCONTROLLER

The Arduino Uno is a microcontroller board based on ATmega328P micro-

controller. This micro-controller board is made of Digital and Analog input/
output (I/O) pins which can be even interconnected with other expansion
boards and electrical circuits. The board contains 14 digital I/O pins (six
for PWM output), 6 analog I/O pins, and which can be programed with the
Arduino IDE, with a type B USB cable. USB cable or any external 9-volt
battery provides power to the Arduino micro-controller board, where the
power is between the range of 7–9 volts. It is similar to Arduino Nano and
Leonardo. The Arduino-Uno board is first in a series of USB-based Arduino
boards. The version 1.0 of the Arduino IDE was one of the reference versions
of Arduino, which are reinvented incorporating various other changes. The
ATmega328 on the board is pre-programed with a bootloader which allows us
in uploading new code without any use of external programmer (Figure 12.5).

FIGURE 12.5 Arduino Uno microcontroller.

12.4.2 NEO 6M GPS MODULE

The NEO-6M GPS module is a well-known name while considering any GPS
receiver which is built with 25 × 25 × 4 mm ceramic antenna, providing a very
strong satellite search capability. The power and signal indicators are very
Microcontroller-Based Wearable Location Tracker System 153

useful, which can monitor the status of the module. It also has the efficient
facility of data backup, in case accidental power failure (Figure 12.6).

FIGURE 12.6 Arduino Uno microcontroller.

12.4.3 SIM800L GPRS/GSM MODULE WITH ANTENNAE

SIM800 is a quad-band GSM/GPRS module that works on frequencies 900

MHz EGSM, 850 MHz GSM, 1,800 MHz, and 1,900 MHz PCS. It is also
featured with GPRS multi-slot class which consists of 12/class 10 (optional),
and also it supports the sequences of CS-1, CS-2, CS-3, and CS-4 GPRS
coding schemes. It also has one UART port. It is also featured with one
USB port which can mostly be used for updating the firmware and useful
for debugging. Receiver output and microphone input are also possible
through audio channels. SIM800L has been configured with one SIM card
interface, which integrates the TCP/IP protocol. The operation of SIM800L
is performed on a supply within the range of 3.4 to 4.4 V. Making calls,
sending, or receiving texts, audio over the internet, are some of the utilities
of SIM800L. Hence, making it possible and useful when applied in projects,
such as home and agriculture automation (Figures 12.7 and 12.8).
154 Advances in Data Science and Computing Technology

FIGURE 12.7 SIM800L GSM module.

FIGURE 12.8 Pin-diagram of SIM800L GSM module.

Microcontroller-Based Wearable Location Tracker System 155

12.4.4 JUMPER WIRES

Jumper wires are simply different colored wires that are used to connect
components without soldering with the help of the pins that are present
at the end of the wires. They are very useful in forming the circuit. And
their different color codes make it easy to differentiate one from the other,
thus easily distinguishable. These are usually used along with breadboards
in order to connect the circuit and change the circuit as and when needed
(Figure 12.9).

FIGURE 12.9 Jumper wires.

12.4.5 BREADBOARD

A tiny holed rectangular box usually made using plastic is generally referred
to as a breadboard. These holes being very useful help one to insert any
electronic components with ease, in the process of making a prototype or a
product. It is a base solemnly used in the procedure of the very basic construc­
tion for making realistic and miniature prototype of an electronic model.
Nowadays, we have breadboards that need no soldering and has perforated
plastic blocks with a number of tin or alloy made clips, and sockets for better
insertion of the electronic components. The clips are many a times referred to
as contact points or tie points. A breadboard can be specified and segregated
with the help of tie points (Figure 12.10).
156 Advances in Data Science and Computing Technology

FIGURE 12.10 Bread board.

12.4.6 SWITCHES

Switch is used to turn the circuit on. The switches that are used in bread­
boards are mostly called, push-button switches. They have four legs to fit
well with the perforations of the breadboard (Figure 12.11).

FIGURE 12.11 Switches.

Microcontroller-Based Wearable Location Tracker System 157

12.4.7 9V BATTERY SOURCE

A 9volt battery source is used a source of power in the circuit. The 9V battery
format is often accessible in its basic carbon-zinc and alkaline chemistry, in
basic lithium iron di-sulfide, and also present in the form of nickel-cadmium,
nickel-metal hydride and lithium-ion, which is rechargeable (Figure 12.12).

FIGURE 12.12 9V battery source.

12.5 METHODOLOGY

The system will be in a form of a wearable device. And each device will
be having a registration key. With each key two mobile numbers will be
registered, in which regular updates of the location of the user will be sent.
The user can change these registered numbers if needed. For changing the
registered numbers, the user will have to log in to a website and make the
necessary changes. Since all the devices have a unique registration key, all
the record of a particular device will be maintained. The information won’t
be disclosed, and will be kept under proper secured supervision.
The user will be able to turn the device on. Once the device is turned
on, the real time location and the time of the user will get logged in the
158 Advances in Data Science and Computing Technology

local database. This information then will be periodically get logged into the
cloud. Similarly, location of the user will be sent to the two registered mobile
number. The location of the user will be sent to the two registered mobile
numbers after a fixed interval of time.
There are many other systems, in the market which gives the location
of the user but there is a specialty of this device. But only tracking of the
location isn’t enough, is it? We are getting the location of the user, but what
if the person is danger? What if the user is unable to make a call, and ask for
help? In that case what is the need of this location tracker? There has to be
something which can give an edge over these simple trackers.
The specialty of this system gives this device an edge over all the existing
devices. It has an exclusive feature of SOS facility. Well, this SOS facility is
not the only specialty. This device can send alerts to the alert center, when
the SOS button is pressed. As soon the alert center receives the alert message
it checks the immediate locality of the distressed user and alerts the nearest
helps available at that point of time, so that the victim, the distressed can be
rescued from that danger. And this unique system makes this system more
special, and the only one of its kind. This system has a lot of battery backup
(Figure 12.13).

FIGURE 12.13 Working principle of this system.

For instance, when a person wearing the device, is in danger, at that time
he/she can press the SOS button, and immediately an alert message will go to
the registered mobile numbers. As this system is unique, as soon as the SOS
button is hit, the recent last location is sent to the immediate local supports
available like fire brigade, police, and ambulance.
Microcontroller-Based Wearable Location Tracker System 159

The system is smart. As the local area changes, the local supports change
too and accordingly this system works. The working of will be clearer with
the next few instances.
For example, a woman wearing the device is walking on the street, and
she is in danger. There are a group of eve teasers following her, or harassing
her. She immediately when presses the SOS button, so an alert message goes
to the alert center. Once the alert center receives this alert message it checks
the present location of the user, and checks for the immediate help available.
The immediate help available locally is the police station, hospitals, so these
places will get an alert message along with the location of the user, and a
rescue operation can be arranged for. Also, another alert message is sent to
the two registered mobile number.
Similarly, when a fisherman loses his boat in the mid sea, and he presses
the SOS button. The SOS message reaches the alert center. The alert center
checks the locality of the user, checking he is the mid sea, the local support
available will be the coast guards; police ambulance in this case will not be
the effective help. Thus, the SOS button on being pressed, the alert message
along with the location of the user is sent to the coast guards, who are the
immediate support avail then (Figures 12.14 and 12.15).

FIGURE 12.14 Pin diagram of Arduino Uno.

160 Advances in Data Science and Computing Technology

FIGURE 12.15 Circuit of the system.

12.6 USED CASES

When the question of application arises, this device has a diverse area of its
significant impact and application:

12.6.1 FIRE FIGHTER SAFETY

A deadly fire broke out in an old multi-story building in an urban busi­

ness area. This house is surrounded other houses and there are thousands
of workers working. All these houses are affected due to this fire. The fire
fighters have started their rescue operation. All of them are trying their best.
The houses being old, the building is on the verge of collapsing. There are
several fire fighters inside the house and all of them are busy rescuing the
victims. In the meanwhile, two of the fire fighters are climbing up a flight
of stairs which is affected by fire. Suddenly the stairs collapse and they are
lying half buried under the debris, calling for help, but nobody is there to
listen to them. Slowly the smoke is choking them to death (Figure 12.16).
After hours of hard and tedious work, the fire fighters finally succeeded
in evacuating the people who were inside the houses. Suddenly, they notice
that two of their brave officers are nowhere to be seen. It is evident that those
Microcontroller-Based Wearable Location Tracker System 161

brave officers have got stuck somewhere inside or have been lying under the
debris. But how will the others know the location of their brave colleague?
This device will help them. This device will be worn by the fire fighters. As
soon as they will enter particular premises, their location will be sent to the
database, and each fire fighter’s location will be tracked. When the fire fighters
face such a situation like the above, they can press the SOS button, and the
alert message will help to locate the victims. Thus, if a fire fighter faces such
a situation like both of them had then with the help of this device, they will be
able to survive. The brave fighters who give fight the fire to save our lives, will
not lose their precious lives due to lack of proper information.

FIGURE 12.16 This device is very beneficial for the fire fighters.

12.6.2 FISHERMEN SAFETY

While we all are in our resting time, at the time fishermen gets ready for their
daily routine. For them only we people are able to have fishes as our meal
but we do not have any idea through how much hard work fishermen goes
for catching those fishes. Every day at the dawn, they prepare themselves for
going at the middle of the sea for fulfilling their duty.
Suddenly, a storm breaks out at the mid shore and the weather condition
around the sea or any other water body drastically changed within a minute and
respectively high tides occurred and the waves are irregular and on tenterhooks
and the fisherman is in the midst of the sea. Can you imagine the condition of
162 Advances in Data Science and Computing Technology

the fisherman? Their life is under threat but we need to assure them a danger
free life. So, our proposed device is a life savior for them. Our proposed device
is of wearable type and fisherman can wear the device and then they can carry
out their work. As soon as they move towards the midst of the sea or any other
water bodies their locations will be recorded and sent to the database and the
each and every fishermen’s location will be tracked.
When fishermen face situation like tempest, they can press the SOS
button and immediately an alert message will be generated and will be sent
to the registered contacts so that immediate rescue operations are arranged.
And if they are not in the condition to press the SOS button, their location
can be tracked and immediate supports could be sent to rescue them. They
risk their lives for us and it is our duty to save and provide them security.

12.6.3 CHILD SAFETY

Nowadays, child safety is a big concern for the urban parents, who mostly
stays away from their children due to work for the entire day almost. At this
situation, the wearable location tracker can be designed and implemented to
be used in their lifestyle. It will definitely prove very effective in ensuring
the safety to their children, even kids. This device will also prove efficient
when the children are out of home, i.e., in schools, coaching classes or in
any other place. The parents are always in a worry of the safety of their
children, where our proposed product can be a boon for them. It is very
easy to monitor for the parents and keep a track of their most loved ones.
It is also very user-friendly and thus, children wearing it, will not face any
sort of problem. It can be used for age group 3 years and above.
As for example, when a child goes to school by a pool car or school
bus, there may be such situations of child trafficking, in that case, the child
just needs to press the SOS button. Immediately, emergency support of
police will be gathered and an alert message will also be sent to the parents’
mobile number which had been registered initially in the device. Thus, it
will take care of the children, even when they are away from their parents.

12.6.4 WOMEN SAFETY

Women are another important part of our society. They are often heard of
becoming victims of bullying and harassment in public. In recent times, the
security of women is a high point of concern for all, even the governing
Microcontroller-Based Wearable Location Tracker System 163

bodies. In this situation, this wearable location tracker can mitigate the
problem up to a certain stage. Nowadays, rape has become a common issue,
which is the call of the hour for us and needs to be put an end to this, as
early as possible. Thus, the women get victimized the most in the society.
For them, the above mentioned device will rescue them, if caught in such
horrendous incidents.
Now, for example, if a woman is traveling alone late night, she defi­
nitely needs security. At this situation, she can wear this device and make
sure that the power button is ON. Thus, in any chance if she faces some
weird or unnatural circumstances, or if she senses danger, can simply hit
the SOS button. As soon as possible, a nearby help, like police can be
provided; and also, an alert message will be delivered to the closed ones
of the victim, and she can be rescued. Thus, it will reduce the cases of
women harassment and will also help in building a better place for them
to live.
Thus, we can see that this device is not only for the fishermen, but also for
the fire fighters, women, children, for each and every section of the society.
This device is for the benefit, for the betterment for the good and security of
the entire society at large.

12.7 RESULTS

The wearable location tracker during disaster, will not only be able to detect
the location of the user, but also will provide support with nearest assistance
in case of any danger suspected. The device will also be helpful to the user
to get rescued by the local supports available. For example, if a person in a
building, caught in dangerous fire breakout, then the assisting body will be
the fire-brigade and police. But, if a fisherman loses his boat in mid water
body or falls in any type of danger, then in this case the rescuing body will be
the coast-guards and not the police. Thus, it is clear from the above instance
that the support or the help that will be generated, entirely depends on the last
tracked location of the user, i.e., the last updated location of the user stored in
the cloud. The person wearing this device will not feel any extra weight, as
it is extremely light in weight, which is one of the key features of the device.
Another striking feature is that, it has an exclusive SOS facility inbuilt in
the device itself, so that at times of any emergency, whenever the button is
hit, local support will be arranged automatically by determining the recent
location of the user. Our device is facilitated with two buttons, first one is the
164 Advances in Data Science and Computing Technology

power button, having the power symbol and green in color and second one is
the most important, the SOS button. Since, it will sense the danger; it has a
vibrant red color to make it more noticeable at times of emergency.
It also has an extra and highlighting facility which proves beneficial
in terms of power saving and also accounts for being more user friendly.
Suppose, the user instead of pressing the power button hit the SOS button or
vice-versa; then in that case, it has the technology that unless and until the
power had not been turned ON, the SOS button will not work and similarly,
if the user hit the SOS button mistakenly instead of power button; then no
alert message will be generated because the device will not start operating
before turning ON the power button. The device is also made water-proof
and even heat-resistant, thus making it fit for use even underwater to assist
the fishermen, whenever he senses danger or loses his boat in sea.

12.8 CONCLUSION

By designing this wearable location tracker, it is ensured that in near future

it can reduce the number of lost cases, also the cases relating to women and
child trafficking, since their location can be tracked using this device. And
not only the location, but also nearest local help is assured to the victim;
and along with this, an alert message will also be delivered to the registered
mobile number of the close ones. Furthermore, the device can also solve
problems related to tracking of fishermen, if caught in some danger in the mid
sea or loses his boat during extreme weather conditions. Hence, reducing the
life risks faced by them. Superior GPS module, GPRS/GSM module are used
which reduces the chances of network failures in most of the remote areas
than expected. In this particular model, Arduino-NANO is used to design
the device, which is better compared to its UNO version based on the size.
Because of using Arduino-NANO, the size of the device is reduced which is
very easy to wear and carry. This location tracker if used efficiently, i.e., if
the SOS button is pressed on time, can really save precious life. It will act as
a savior to the human being.

12.9 FUTURE SCOPE

In future, this location tracker can be given a new shape with some added
features, which will make this more efficient. We are continuously working
to increase the efficiency of our product by applying various available latest
Microcontroller-Based Wearable Location Tracker System 165

technologies. In this device, Arduino-Uno microcontroller has been used, but

in near future Flora micro-controller can also be used to scale it down to a
much smaller size than present. The wearable location tracker, is formulated
with a SOS button, which needs to be hit at times of emergency; but in future
this can also be made fully automatic, i.e., the device will not require any
sort of human intervention to press the SOS button during emergency. The
button will be then automatically activated and start performing the needful
by determining the vital body parameters and also some advanced technolo­
gies which will determine the type of hazard the person is in. This device can
also be applied in cases of child safety and women security in recent times
and in near future. It can also find its application in the field of defense,
by serving our country’s backbone, our defense sector. Thus, serving the
mankind at its best.

12.10 BUSINESS PLAN

This device will not only secure a lot of safety, but also has a potential to
give a way of living to a lot of people and secure their lives. This device will
be manufactured by a particular company and the business setup will ensure
monopoly regarding the selling, maintenance of the service. This company
will also deal with the other aspects related to the device such as security of
information and the data hiding part. This company will also have provisions
for the emergency alert systems, which will be very active and quick to
response when there is a time of emergency.
As mentioned earlier each device will be registered one, so the entire
data will be logged into the company’s log, where with each device corre­
sponding details of the user will be maintained. Since the company will
have the policy of monopoly, there will a lesser threat of duplicate products.
Moreover, there will be no compromise with the quality of the device and
especially the safety and security of our users.
As mentioned above, the information will not be disclosed, and will be
kept under proper secured supervision, for that special and efficient team
will be present, who will look into the matter of safety and security of the
data, and will be supervising the entire event.
All these work will need a lot of people to work together. There will be
offices of this company in various parts of the world. Thus, this company
will be a home to thousands of employees. This company will bring smiles in
166 Advances in Data Science and Computing Technology

the faces of millions of people by offering them jobs, and ensuring the safety
and security of the users.
Each device will cost approximately Rs. 2,250/-. And as the number of
production increases, and the production is done on a large scale, there will
be a considerable drop in the price of each device.

KEYWORDS

• Arduino microcontroller
• cloud
• global positioning system (GPS) module
• global system for mobile communications (GSM) module
• SOS button

REFERENCES

An end-to-end location and regression tracker with attention-based fused features. In: 2019
International Joint Conference on Neural Networks (IJCNN)Date of Conference: 14–19
July 2019. Date Added to IEEE Xplore: 30 September 2019 INSPEC Accession Number:
19028209. doi: 10.1109/IJCNN.2019.8852288Publisher: IEEE.
Debajyoti Pal, Anuchart Tassanaviboon, Chonlameth Arpnikanondt, & Borworn Papasratorn
(2020). Quality of Experience of Smart-Wearables: From Fitness-Bands to Smartwatches
(Vol. 9, No. 1, pp. 49–53). In: IEEE Consumer Electronics Magazine Date of Publication: 04
December 2019 INSPEC Accession Number: 19194203. doi: 10.1109/MCE.2019.2941462.
Publisher: IEEE.
Design and development of wearable GPS tracking device by applying the design for wearability
approach. In: 2018 4th International Conference on Science and Technology (ICST), Date
of Conference: 7–8 Aug. 2018. Date Added to IEEE Xplore: 12 November 2018 INSPEC
Accession Number: 18233756. doi: 10.1109/ICSTC.2018.8528636 Publisher: IEEE.
Design of handheld positioning tracker based on GPS/GSM. In: 2017 IEEE 3rd Information
Technology and Mechatronics Engineering Conference (ITOEC), Date of Conference: 3–5
Oct. 2017. Date Added to IEEE Xplore: 01 December 2017INSPEC Accession Number:
17411810. doi: 10.1109/ITOEC.2017.8122477 Publisher: IEEE.
Development of real time life secure and tracking system for swimmers. In: 2019 3rd
International Conference on Computing Methodologies and Communication (ICCMC), Date
of Conference: 27–29 March 2019. Date Added to IEEE Xplore: 29 August 2019INSPEC
Accession Number: 18958280. doi: 10.1109/ICCMC.2019.8819859 Publisher: IEEE.
Development of real time life secure and tracking system for swimmers. In: 2019 3rd
International Conference on Computing Methodologies and Communication (ICCMC), Date
Microcontroller-Based Wearable Location Tracker System 167

of Conference: 27–29 March 2019. Date Added to IEEE Xplore: 29 August 2019INSPEC
Accession Number: 18958280. doi: 10.1109/ICCMC.2019.8819859 Publisher: IEEE.
Energy efficient assisted GPS measurement and path reconstruction for people tracking.
In: 2010 IEEE Global Telecommunications Conference GLOBECOM 2010, Date of
Conference: 6–10 Dec. 2010. Date added to IEEE Xplore: 10 January 2011INSPEC
Accession Number: 11743331. doi: 10.1109/GLOCOM.2010.5684306 Publisher: IEEE.
Foxlin, E. (2005). Pedestrian tracking with shoe-mounted inertial sensors. In: IEEE Computer
Graphics and Applications (Vol. 25, No. 6, pp. 38–46). Date of Publication: 07 November
2005. INSPEC Accession Number: 8662444. doi: 10.1109/MCG.2005.140. Publisher: IEEE.
Instantaneous feedback pedometer with emergency GPS tracker. In: 2018 2nd International
Conference on I-SMAC (IoT in Social, Mobile, Analytics and Cloud) (I-SMAC) I-SMAC (IoT
in Social, Mobile, Analytics and Cloud) (I-SMAC), 2018 2nd International Conference on
Date of Conference: 30–31 Aug. 2018. Date Added to IEEE Xplore: 28 February 2019INSPEC
Accession Number: 18490586. doi: 10.1109/I-SMAC.2018.8653718 Publisher: IEEE.
Internet of Things: Solar array tracker. In: 2017 IEEE 60th International Midwest Symposium
on Circuits and Systems (MWSCAS), Date of Conference: 6–9 Aug. 2017. Date Added to
IEEE Xplore: 02 October 2017Electronic ISSN: 1558–3899INSPEC Accession Number:
17214744. doi: 10.1109/MWSCAS.2017.8053109Publisher: IEEE.
Optimization of laser trackers locations for position measurement. In: 2018 IEEE International
Instrumentation and Measurement Technology Conference (I2MTC), Date of Conference:
14–17 May 2018. Date Added to IEEE Xplore: 12 July 2018 INSPEC Accession Number:
17917295. doi: 10.1109/I2MTC.2018.8409835 Publisher: IEEE.
User localization using wearable electromagnetic tracker and orientation sensor. In: 2006 10th
IEEE International Symposium on Wearable Computers Date of Conference: 11–14 Oct.
2006. Date Added to IEEE Xplore: 22 January 2007Print ISBN: 1–4244–0597–1 INSPEC
Accession Number: 10236845. doi: 10.1109/ISWC.2006.286343Publisher: IEEE.
CHAPTER 13

DESIGN AND MODIFICATIONS OF

EYEDROP BOTTLE HOLDER DEVICE:
A SIMPLEST WAY TO INSTALL OCULAR
DRUG INDEPENDENTLY
SIRAJUM MONIRA, VISHAL BISWAS, and SATYAM JHA
Department of Bachelor in Optometry, Pailan College of Management
and Technology, Joka, Kolkata–700104, West Bengal, India,
E-mail: [email protected] (Sirajum Monira)

ABSTRACT

To design a simple eye, drop bottle holder device that can be used to put
ocular drugs independently even the patients with severe visual impairment.
We have worked in three phases; in Phase 1, designing of the non-electronic
device (eye dropper) it is a non-electronic device and we are calling it “Easy
eye dropper.” It involves less effort for the patient. The leverage which is
present, it is specially designed with extended arms, so that it can deliver eye
drops using with the lightest finger press and drop will be fallen inside the
eyes. The eyepiece and locks position is constructed in such way that is ideal
for the user. Now as the size of the drops varies from length to diameter (5
ml to 10 ml), all the diameter and length of all the drops are not same which
are available in the market, to get rid of it we a have a way out here (as we
used the screw tightened mechanism of a compass and a pencil). Like the
same way, here we have placed a screw system, where you can put the eye
drop inside the device and then according to the diameter you just tighten the
screw. Hence it will be fit inside, then place the device on the eye, by the help
170 Advances in Data Science and Computing Technology

of leverage press it automatically, with one gentle press a drop will come out
and will go directly inside the eye.
In Phase 2, designing of the electronic device (electronic eye dropper).
We have designed another electronic model with some additional features
which can be considered as the upgraded version of the non-electronic
model. It includes an eye detector sensor and some audio output along with a
micro vibrator. This technology will reduce the effort of the patients in terms
of his/her attentiveness as with this electronic technology a sensor will detect
that whether the tip of the bottle is placed on its proper place or not. A beep
electrical sound will confirm the number of drops fallen in the eye. Also, we
have incorporated a micro vibrator in it keeping the deaf and mute patients
who may not detect the sound sensor.
One press = one beep sound = one drop will get out

In Phase 3, practical approach (pilot project) we have a plan to conduct

a pilot project with this newly invented device to see the durability of one
eye drop bottle manually and by these electronic and non-electronic devices.
The construction of “Easy eye dropper” is very simple in use for all the
subjects whenever they try to install the eye drops in their eyes without
anyone’s help. It also saves the number of attempts required to install the
eye drops but the effectiveness of this device yet not tried practically. This
phase will be done in the coming days. A new modified eye drop bottle
holder may be an effective device and useful invention in terms of saving
the time, effort, wastage of drug quantity and will create independence of
self-installing drugs into eyes.

13.1 INTRODUCTION

There are a majority of ocular diseases where the mainstay of treatment of the
patients includes topical therapy. If we overview the various ocular conditions,
starting from a very minor ocular disturbance to a severe chronic problem is
treated with the help of topical therapy along with its other additional treat­
ment. As there are many kinds of eye drops, it is hard to generalize about
their usage area, but the more commonly the areas we are always dependent
on topical drugs mostly: Dilating drops during eye examinations, lubricating
drops for dry eye symptoms, redness-relieving drops, anti-allergy drops for
itchiness-relieving, pre-operative, and post-operative drugs, commonly, and
frequently used antibiotic drugs and the pressure-lowering drugs used for
Design and Modifications of Eyedrop Bottle Holder Device 171

long term treatment like glaucoma. Consider a disease like glaucoma where
we have to make sure to put in the right amount of eye drops at the right time.
This way the drug will help to keep the eye pressure at the correct level for the
patients, which decrease the risk of their eyesight becoming poor in future.
Many people find it difficult to put eye drops in initially, also many of the
patients cannot even put drugs independently and it is easy to miss the regular
prescribed doses. Glaucoma is a slow progressive ocular condition that can
ultimately result in impaired vision or even person may reach to the condition
of blindness, so it is important to take all the drops as prescribed. Also, for
glaucoma sufferers, most of them have to use drops even for their lifetime.
There are various ways to install eye drops in the eye and many people
take many ways for this task. Some patients try to put independently but
some people take others’ assistance. Before to start putting a drop one has
to make sure that the hands are properly cleaned. Few drops need to be
shaken. Remember. One of the simplest ways of installing drops in is to sit
comfortably in front of a mirror or can be instilled drops whilst lying down,
need to gently pull down the lower eyelid with a finger of any hand, with
the other hand which is free, need to press the bottle according to the proper
instructions and allow the drop fall into the space between eye and the lower
lid. After putting the drop in the eye, the patient needs to close the eye lid
gently and press on the inside corner of the eye, near the nasal area, with a
finger for one minutes. This will help to reduce the rate at which the drop
drains out through the tear duct into the throat cavity, rather than staying
in the eye where it is essential. A little amount may still drain through the
tear duct and be swallowed in our throat, that is why it felt bitter taste in
throat. This small amount which we are swallowing is not harmful but, the
amount of eye drops entering body through the tear duct is minimized. This
is why there is a higher chance of unnecessary wastage of drugs and also the
absorption of drug through the ocular surface reduced. Many times, patients
also get suspected whether the right quantity of drugs has gone inside the eye
or not, hence they put another drop of the drug from this suspicion which
may lead to get the bottle finishes early. For this reason, most of the topical
drug user takes someone’s help always. To be dependent always for this task
on family members, friends or other people may reduce the wastage of drugs
but it leads to untimely putting of drugs due to the unavailability of helping
hands on time always. If it is a glaucoma patients or post-operative cases
then there is always a need for frequent eye drop installation to preserve the
eyesight. There are some other situations also where one has to install eye
drops independently. Mainly, the time when people are there in their work
field or traveling alone in outdoors.
172 Advances in Data Science and Computing Technology

Keeping all those points in mind we have made “Easy Eye Dropper”
which may help to get rid of both the problems in terms of drug wastage and
also a dependency on drug installation. There are few aids already commer­
cially available for this task but they have a few disadvantages which are
leading to its non-popularity among the ocular drug users. Mostly all these
aids are manual or non-electronics. Few designs are not suitable for all sizes
of bottles, few aids not proper for gripping in hands and some are creating
problem in positioning proper places.
New modified eye drop bottle holder “Easy Eye Dropper” will be an
effective device and useful invention in terms of the longevity of drug quan­
tity, wastage of drug quantity and will create independence of putting the
drug into eyes and will overcome all the limitations of previously made aids.

13.2 OBJECTIVES

The main aim of our work is to design a better technology eye drop bottle
holder device. Our goal is to help all the ocular drug users including the
severely visually impaired people, hand tremors, arthritis, and old-aged who
cannot put eye drops without any ones help and always missed the drug
schedule due to lack of helping hands on time. Also, this technology may
reduce the wastage of drugs which may indirectly save the health budget of
any drug user. The specific features of lightweight, easy usage technique and
low cost gives ocular drug user a new lease of life, a new dimension of inde­
pendence and enables him to save some budget on ocular drugs purchasing.

13.3 REVIEW OF LITERATURE

Taneja et al. group published a paper on Innovative eye drop applicator

namely bull’s eye applicator. This applicator is helpful for self-installing of
eye drops independently. They reported the effectiveness of this eye drop
device in preventing the wastage of medicine. In this device here they have
invented a small pocket sized device where there is an easy handy attachment
which can be attached to any kind of eye drop bottle. They also performed a
study where they got a result that the use of this device reduced the wastage
of drugs. The dispenser was used in few subjects divided into two groups;
they were asked to apply the lubricating eye drops of (specific size bottle)
in any one eye between the two without the help of device and in fellow eye
with the help of this device. A count was kept on the attempts made by the
Design and Modifications of Eyedrop Bottle Holder Device 173

users for the application of eye drops and the residual eye drops which were
not used in the returned bottles were measured for quantitative assessment as
planned in the objective of their study.
They concluded that the number of attempts of instillation which was
made with the help of eye drop applicator was reduced significantly. This
device also reduced the wastage of drugs.
But they found few limitations also like severe arthritis patients and low
vision patients cannot use this. The mirror which is placed in the applicator
can break easily (Mukesh, Chappidi, and Naga, 2016).
Jungueira et al. made a study on the efficacy of drugs and also to check
the safety of a new device (eye drop R), which had been designed for eye
drop instillation in patients having or not having glaucoma.
They have done this study on both types of patients (on glaucoma
patients and healthy participants) on applying the eye drop. This study
included patients with glaucoma and normal healthy participants after doing
all comprehensive eye examination and measurement of the baseline intra­
ocular pressure, the topical hypotensive medication had been given in both
eyes and the Eye drop delivery device was made available to all subjects
(with video and written instructions in them for proper evaluation) for use
in one eye, which was randomly chosen by the participants. In the second
phase, all the participants were evaluated by an examiner for IOP, inves­
tigation of possible related side effects, and the ease of putting eye drops.
Conclusion made was that Eye drop gained a better subjective response
regarding the ease of installation as compared to traditional instillation.
But this device also has some limitations like it is an invasive technique:
which can cause harm to the ocular surface. It does not give a proper idea that
how much amount of drop got instill (Junqueira, Lopes, and Souza, 2015).
George C. Bauer invented a device in the year 1992 by which eye drop
applications became easy. The Eye drop applicator comes for an eye dropper
vial comprising of a cup with a lip portion to get fitted well around the eyeball
and helps in letting the eye lid remain opened. A sleeve has an internally
threaded passage extending from the top of the cup. The internally threaded
passageway can be removed and it is connected to an externally threaded
neck of the eye dropper vial. The one who uses it will have to squeeze the
eye dropper vial for directing an eye drop from the nozzle of the eye dropper
vial onto the eyeball center.
This applicator also has some disadvantages like this was just a theory
no working model was produced according to this theory. If it would have
produced then this device would have been an invasive device which could
174 Advances in Data Science and Computing Technology

have caused harm to the eye during instillation of drug (www.maddak.com/

autodrop-eye-drop-guide-pr-28115.html?page=2).
Shai Recanati invented another type of eye drop applicator device in the
year 2004. Here they invented an eye-drop applicator for putting eye-drop into
a user’s eye without the blinking one’s eye. Eye drop applicator is inclusive of
a container which has a neck and also has an opening end for storing the solu­
tion; it has a lid too; and further includes a tubular eyedropper member being
attached to the lid and through which eye-drops come out from the container.
This device also has limitations like the above one (https://www.guldenoph­
thalmics.com/products/index.php/opticare-eye-drop-dispenser.html).
Another applicator called auto drop which was manufactured by the
British Royal National Institute for the Blind. There is an arrangement,
which holds the eye open while keeping the bottle over the eye and also it
consists of a very small pinpoint like hole that allows light to enter in; as per
the manufacturer, when the patient looks at the pin hole, the eye is rightly
positioned to receive the drop and drop does not get waste while putting. But
this device is also not suitable for the severe arthritis patients as it involves
the patient’s own press for instilling the drop. Also, this device is not at all
available in the Indian market (https://magic-touch-eye.myshopify.com/).
Cameron Graham made an eye drop dispenser which is called Opticare Eye
drop dispenser. This device is designed in such a way that it make easier for the
patient to press the bottle. But there is no such studies done till yet, which tells
us the effectiveness of this dispenser among the user. Also, this device is not at
all available in the Indian market (https://www.contactlenskit.com/optiaide/).
Another technology of installing eye drop is by using capillary action
technique. It involves putting a single drop into a holder and then bringing
the holder very close to the eye, allowing the drop to fall onto the eye from the
device. This technique had several advantages, including increasing the chances
that only one drop will be used at a time and no more drops will be wasted;
no need for the patient to tilt his head back during the installation procedure
(https://www.amazon.com/Cress-Dropper-EyeDrops-Professional-Model/
dp/B004R8TOGM).
The Magic Touch—The Eye Drop Helper is invented by Julius Shulman.
This device, has a small rubber bell about the size of a thimble, is placed on
a fingertip; with the finger positioned vertically, a drop is placed into a small
groove at the tip of the device. The user then moves forward, and when the
device gets close to the user eye, the drop falls onto the eye. But it has some
drawbacks also like always needs to keep the device clean as it is an invasive
technique. Low vision patients cannot use it. There is no study regarding the
effectiveness of this procedure on the patients (Sharma et al., 2016).
Design and Modifications of Eyedrop Bottle Holder Device 175

OptiAid is another type of device which looks more high-tech than the
other available devices, it can hold up to five drops in its five chambers.
Unlike the other devices, this one is positioned in front of the eye and
allows the drop to fall onto the eye, theoretically preventing the device from
touching the cornea. However, it needs moving the eyelids and eye lashes
out of its way to avoid the touch. Negative factors are: Not available in the
Indian market. It is a complete invasive technique to install drug. No studies
have still done on this innovation regarding its effictiveness (Efron, 2013).
Another aid created by Jonathan Cress an ophthalmologist. This dropper
is an arched tube. It marks the users nose as a reference point so the tip can be
steadily placed in the medial canthi area and the area will be properly identified.
Once the tip is properly placed in its position, the bottle is squeezed; the
drop leaves the aperture, slides down properly and off the little ramp onto the
delivery tip and into the user eye. The user releases the bottle when they feel the
drop fallen into the eye. The current version of the device is with no measuring
component, but mostly the patients gain an understanding of squeezing the bottle
to release just one drop of medicine and extra drops will not be wasted. Still this
device also has some limitations. Patients who have weak hands find it difficult
or impossible to press the bottle by squeezing accurately (Kahook, 2007).

13.4 METHODOLOGY

The study was carried out into three phases. But in this chapter, we have
described only two phases and the third phase will be done in our next work.
Broadly, phase I involved in the design and implementation of the eye drop
bottle holders which will be working manually. We called the aids “Easy Eye
Dropper”(non-electronic model).
In phase II, we used Arduino and Ultrasonic based sensor techniques.
This is an electronic model of “Easy Eye Dropper.”
In phase III we have to assess the practical and functional approach of
the newly made “Easy Eye Dropper” aids with both the electronic and non­
electronic models.

13.4.1 PHASE-I (EASY EYE DROPPER: NON-ELECTRONIC MODEL)

This model is beneficial particularly for those who have severe arthritis,
spondylitis (mainly cervical and for that they cannot move back their neck
to the backside) or difficulty in lifting their hand up and reach close to the
176 Advances in Data Science and Computing Technology

eye due to any type of systemic illness. This type of people mainly faces
a lot of challenges while they are trying to install eye drops in their eyes
independently. Hence, they have to be always dependable on others. The
leverage that one can get from this specially designed eye dropper holder–
two extended arms (14 cms) which to deliver eye drops just by using even
the gentle finger touch. The eye piece arrangements and locks in the position
that is perfect for you during the drug installation procedure.
Here we have implemented a screw in the dropper – now as the size of the
eye drop bottles varies from length to diameter based on its commercial compa­
nies, along with the quantity of drugs also varies from company to company
and brand to brand, e.g., 2.5 ml–10 ml, diameter varies from 3 cm–3.5 cm. The
shape of the eye drop bottle also varies from round to oval (Table 13.1).

TABLE 13.1 Explaining the Procedure of Fitting of Bottle Into “Easy Eye Dropper” and a
Simple Instructions for the Users
Fitting of Bottle Into “Easy Eyedropper” Instructions for the Users
If the bottle is new remove the protective seal Open eyepiece and remove the bottle cap.
(Figure 13.1(a)) Tight the screw or lose the screw according
to the size of the bottle (Figure 13.2(a))
Open eye piece towards arms/leverage of the Close the eyepiece. Rotate the eye piece
dropper (Figure 13.1(b)) towards left or right as per in which eye
you are using for the most comfortable
administration of the drug (Figure 13.2(b))
Place the dispenser on the flat surface and While sitting back or lying down, the user
spread the two arms of the dispenser open, has to tilt his head back and place eye piece
push eye drop bottle inside dispenser until the securely over the eye.
base of the eye drop is fully depressed. Eye piece has finger space for pulling down
Tight/lose the screw according to the size/ lower eyelid with free hand (Figure 13.2(b))
diameter of the eye drop bottle (Figure 13.1(c))
Then close the eyepiece Gently press the arms of device to install
eye drops one per squeeze (Figure 13.2(c))

So, to get rid of these we have a way out here (as we use the compass
during geometry in mathematics, we tightened the pencil in the compass in
the other arms of the compass for drawing) like the same way we have a
screw system in this device. The user needs to put the eye drop bottle inside
the dropper and then according to the diameter one can tighten the screw, this
way the bottle will be fit inside. After that the device needs to be placed over
the eye, by the help of the leverage the extended arms need to be pressed and
automatically by this gentle pressure a drop will come out and will be fallen
inside the eye.
Design and Modifications of Eyedrop Bottle Holder Device 177

(a) (b) (c)

FIGURE 13.1 (a) If the bottle is new remove the protective seal; (b) open eye piece towards
arms/leverage of the dropper; (c) place the device on the surface and spread arms open, push
the eye drop bottle into the device until base is fully depressed into it tight/lose the screw
according to the size/diameter of the eye drop bottle.

(a) (b) (c)

FIGURE 13.2 (a) Open eyepiece and remove bottle cap. Tight the screw or lose the screw
according to the size of the bottle; (b) close the eye piece. Rotate the eyepiece towards left or
right as per in which eye you are using for the most comfortable administration of the drug;
(c) gently press the arms of device to install eye drops one per squeeze.

13.4.2 PHASE-II (EASY EYE DROPPER: ELECTRONIC MODEL)

The steps involved in mechanism of “easy eye dropper” are as follows:

• The device has to be placed on the ocular surface at first, in such a
position so that the drop falls on the eye itself;
• Then the user has to Switch on the device;
• The IR sensor which is present beneath the device will detect the
ocular surface and it will send the message to the Arduino. Then the
Arduino will send this message to the motor to create torque and to
start the working of the lever;
• Out of the two levers one will be fixed at a side another will press the
drop bottle by creating force onto;
• Then by One press by the lever = one beep sound will be coming
out = which will indicate that one drop of medicine from the bottle
has come out.
178 Advances in Data Science and Computing Technology

The electronic “Easy eye dropper” consists of five main major compo­
nents. They are as in subsections.

13.4.2.1 INFRARED (IR) SENSOR

An infrared (IR) sensor mainly made up of two components: an IR light

emitting diode (LED) LED and an IR photodiode. Even though an IR LED
looks like a regular LED, but the radiation emitted by IR LED is not visible
to the human eye. Since the IR sensors can catch hold of the radiation emitted
from the IR transmitter, are even called IR receivers. IR receivers come two
forms that is called photodiodes and phototransistors. IR photodiodes are
different from the normal photo diodes, as the IR photodiodes detects only
IR radiations. Post the IR transmitter emits radiation, a part reaches in the
object and some of it gets reflected to the receiver. Based on the intensity by
the IR receiver, the output of the sensor is defined (Figure 13.3).

FIGURE 13.3 (a) IR sensor which we have used in our study; (b) diagram of the IR sensor.

Specifications:
• Operating Voltage Range: 3.0 V–5.5 V;
• Device Detection Range: 2 cm–35 cm;
• Electricity Consumption: At 3.0 V: ~23 mA, at 5.0 V: ~43 mA;
• Output Level: Output level will reduce /stop on detection of location;
• Location Detection: LED indicator.
For the connection of IR obstacle sensor to Arduino Nano, the hardware
and software needed are:
• IR obstacle sensor module;
• Arduino nano;
• Arduino IDE (1.0.6 V).
Design and Modifications of Eyedrop Bottle Holder Device 179

Hardware connections to the sensor:

• Vcc to 5 V;
• Gnd to Gnd;
• Out to digital pin 7.

intIRsenseor = 7;

int buzzer = 8;

int lever=4;

void setup()

pinMode(IRsenseor, INPUT);

pinMode(lever, OUTPUT);

pinMode(buzzer, OUTPUT);

Serial.begin(9600);

void loop()

int a =digitalRead(IRsensor);

serial.print(a);

if(a==0)

digitalwrite(lever,HIGH);

digitalWrite(buzzer,HIGH);

delay(500);

else

digitalWrite(lever,LOW);

digitalWrite(buzzer,LOW);

delay(500);

Program for IR Sensor

180 Advances in Data Science and Computing Technology

13.4.2.2 ARDUINO

Arduino nano board is used as the main part of the electronic eye dropper.
The Arduino can control the levers by receiving input signals from the IR
sensor. The Arduino boards are arranged with sets of digital and analog
input/output (I/O) pins that are connected one with the IR sensor, one with
the buzzer and one with the lever (Figure 13.4).

FIGURE 13.4 Arduino Nano used in our study.

13.4.2.3 SOUND COMPONENT

1. Buzzer: Here we have included a buzzer just because to make the

patient aware of that how much drop they have to install on their eye.
It helps the patient/the user to get to know that how much amount of
drop went off from the drop bottle. A dropper is operated by trans­
ducer and it converts electric, oscillating signal in the audible range
of 20 Hz to 20 KHz (Figures 13.5 and 13.6).

13.4.2.4 POWER REGULATORS

The input voltage to the Arduino board has been used with 9 volts battery
regulated power source (Figure 13.6).
Design and Modifications of Eyedrop Bottle Holder Device 181

FIGURE 13.5 Buzzer.

FIGURE 13.6 Battery.

13.4.2.5 DC MOTOR

DC motor is also an important part of this project, the main mechanism is

mostly dependent on the DC motor, once the IR sensor detects the surface
it sends the signal to the Arduino sends the signal to the DC motor, after
receiving the signal from the Arduino the DC motors rotates with a speed
of 3000 RPM and thus activating the levers. So, from one end the lever will
press the drop bottle and from the other end the lever will stay static as a
182 Advances in Data Science and Computing Technology

result a force will be created and hence a drop from the bottle will come out.
The DC motor will rotate and a small clip will rotate with it which will go
and hit a string which is attached with a spring by which the lever will get
activate. It will follow the mechanism of cam follower in Creo elements with
compression spring with motion (Figure 13.7).

FIGURE 13.7 DC motor used in our device.

13.4.6 PHASE III: FUNCTIONAL APPROACH OF EASY EYE

DROPPER

This phase has not been explained in this chapter as this phase will be carried
out later. But we have done the planning for this phase. Phase III will be
carried out at Pailan Eye Clinic, Kolkata.
We are going to include 30 subjects of various categories (10=glaucoma,
10=Post-Op cataract and 10= Normal Individuals or volunteer). Those
subjects will be asked to do an exercise for a period of two months. In the
first month they will be asked to use a bottle of approx. 5 ml drug in one eye
without the help of any kind of device in the right eye and in left eye to put
the drug with the help of non-electronic “Easy Eye dropper” model. Next
month they will be given both the non-electronic and electronic model of
“Easy Eye dropper” devices and they will be instructed again to put the eye
drop with the help of the non-electronic model in right eye and electronic
Design and Modifications of Eyedrop Bottle Holder Device 183

model in left eye. Both the time they have to be use two different bottle for
two eyes. After each month’s trial or when the drops in one of the bottle
finished, the users will be asked to return the used bottle. The amount of
the residual drugs which is not been used in the returned bottles will be
measured with the help of a measurement syringe.
Also, we are going to ask few general questions to get the feedback on
this newly made aid. All procedures and protocols will be standardized. This
will be ensured by proper instructional and handy training to the subjects
before to hand over the devices. Examiners also will be instructed properly
about the assessment procedure for those who will be involved in this study.
A Single blinded technique has to be followed during the assessment of feed­
back round. This part of this study is approved by the institutional review
board (IRB) and informed written consent will be taken from all the included
participants.
Three parameters which will be mainly observing are: (i) total times of
applications without any device and with both non-electronic and electronic
devices; (ii) total number of attempts; (iii) wastage of drugs.

13.5 RESULTS AND DISCUSSION

It is worth mentioning that at this point the aim of our study fulfilled as per
our objectives. We started our work with three objectives; the first one is to
design and implementation of a non-electronic eye dropper aids. We have
designed this model with all the possible features of lightweight, handy, easy
to use and cheaper cost. These features made this device more unique as
compared to all other existing aids which are commercially available in the
market. Most of the existing devices are not suitable for patients with severe
arthritis or those who have weaker hand grip so that it is very difficult for
them to hold those devices properly and to locate exactly over the eye where
exactly they need to put the drop. But our design has the advantage of placing
it exactly over the eye because of the design of the holder of this device. For
the same reason other devices are not equally helpful for the patients with
severe visual impairment. Some of the existing devices are fitted with mirror
and users have to locate the tip of the bottle on the eye by looking in the
fitted mirror, but again this would create some difficulties for the users who
have very severe visual impairment. Another disadvantage is the mirror may
break easily. In our non-electronic design, we do not have any such kind of
mirror arrangement which enhances the durability of this aid.
184 Advances in Data Science and Computing Technology

Few devices designed in such a way that the tip of the bottle needs
to touch the ocular surface. This is very harmful in terms of the invasive
procedure as well as it can be the source of infections. Another way, the
invasive technique may damage the ocular surface also if the user get a little
of bit distraction of mind or handshake while installing the drugs. Another
disadvantage of the invasive technique is that the user needs to keep the
device clean as it may contaminate because of the multiple touches. In our
non-electronic device, there is no chance of touching the ocular surfaces
because of the design of its holder.
Though most of the devices are commercially available but still very few
designs are available in the Indian market. So, most ocular drug user are still
unaware of these small technique, hence they are either dependent on others
or wasting eye drops while they are trying to put independently.
The phase two design is completely a newer invention. There are no such
electronic aids for this task. Our main aim of making this electron design was
to remodeling or modifies few features which may add few extra benefits for
the users. In non-electronic device the tip of the device has to put manually
over the proper location so there is a still chance of missing the proper eye
location where exactly one needs to put the drug. The electronic design can
detect the exact location with the help of IR sensor which is connected with
an Arduino sensor. The sensor will detect the eye position and once detected
it will confirm through a sound that will be coming from a buzzer.
Also, the buzzer will create a sound to make the patient aware of that
how much drop they have to install on their eye. It helps the patient /the
user to get to know that how much amount of drop went off from the drop
bottle. These unique features will definitely reduce the number of trials while
putting eye drop and equally will reduce the wastage of drugs as compared
to any existing non-electronic aids.
We are going to do few additional things in the phase III part of our study.
This is completely a new approach that has never been done in most of the
previous works. The previous researches show only the constructions and
implementations of a non-electronic eye dropper aid with multiple features
but most of the studies have not shown the results in real-world situations.
So, there will be a bit doubt for the user about the effectiveness of these kinds
of aids in a real world situations and how it is operated by the actual needy
patients mainly glaucoma, post-operative cases or a severe visual impairment
who are the frequent ocular drug users. The results from phase III must be a
positive indicator to differentiate among the without device, non-electronic,
and electronic one. However, still in this chapter, we have shown only the
Design and Modifications of Eyedrop Bottle Holder Device 185

design and construction phases. Once we will complete the phase III, we
will be adding extra information to prove the effectiveness of our invention.

13.6 CONCLUSION

The non-electronic and electronic “Easy Eye dropper” design will definitely
add an extra platform for the frequent ocular drug users who are hindering
the barriers of putting drugs in day-to-day basis. It leads to good results for
all kinds of users including glaucoma, post-operative cases, or nay general
drug user. This is reliable, light weight, portable, low power consumption,
and low cost. The effectiveness and affordability will make it more accept­
able to the users. This new improvisation in this technique will surely add
an extra benefit in terms of time saving, eye drop installing effort and cost
effective in the process of independent-installation the eye drops.

KEYWORDS

• easy eye dropper

• electronic Arduino
• motor
• non-electronic
• power regulators
• sound component

REFERENCES

Efron, N., (2013). Putting Vital Stains in Context, 96(4), 400–421.

https://www.amazon.com/Cress-Dropper-EyeDrops-Professional-Model/dp/B004R8TOGM
(accessed on 08 December 2021).
https://www.guldenophthalmics.com/products/index.php/opticare-eye-drop-dispenser.html
(accessed on 08 December 2021).
Junqueira, D., Lopes, F., & Souza, F., (2015). Evaluation of the Efficacy and Safety of a New
Device for Eye Drops Instillation in Patients with Glaucoma, 9, 367–371.
Kahook, M. Y., (2007). Developments in Dosing Aids and Adherence Devices for Glaucoma
Therapy: Current and Future Perspectives, 4(2), 261–266.
186 Advances in Data Science and Computing Technology

Mukesh, T., Chappidi, K., & Naga, S., (2016). Innovative Bulls Eye Drop Applicator for Self-
Instillation of Eye Drops, 43(3), 11.
Sharma, R., Singhal, D., Shashni, A., Agarwal, E., Wadhwani, M., & Dada, T., (2016).
Comparison Of Eye Drop Instillation Before and After Use of Drop Application Strips in
Glaucoma Patients on Chronic Topical Therapy, 25(4), e438–440.
www.maddak.com/autodrop-eye-drop-guide-pr-28115.html?page=2 (accessed on 08 December
2021).
https://magic-touch-eye.myshopify.com/ (accessed on 08 December 2021).
https://www.contactlenskit.com/optiaide/ (accessed on 08 December 2021).
CHAPTER 14

AGRO-NUTRITION ALERT (ANA)

RAJNEESH KUMAR,1 PRIYA ANAND,2 DEBAPARNA SENGUPTA,1 and
MONALISA DATTA1
1
Department of Electrical Engineering, Techno International New Town,

Kolkata–700156, West Bengal, India,

E-mail: [email protected] (R. Kumar)

2
Department of Information Science and Engineering, AMC

Engineering College, Bangalore–560083, Karnataka, India

ABSTRACT

This chapter is in regard of our major project named agro-nutrition alert

(ANA). As most of our group members are from a family with agricultural
backgrounds, we have seen farmers facing the issue of low fertility of soil due
to mismanagement of chemical fertilizers. Everyone knows that evolution of
chemical fertilizers was a huge blessing for farmers in India. Farmers were
using chemicals blindly, according to their wish and understandings. But they
realized that it was a “temporary fertility.” Excessive use of these fertilizers
turned the crop-fields barren. During floods, these fertilizers polluted
the water also. The life in ponds got destroyed following several other
consequences. After facing such critical issues, in recent years, government
has arranged a chemical laboratory in each district (Note: Each district has
minimum of 1,000–1,500 villages under its jurisdiction) for execution of soil
test. But it is a time-consuming process with huge economic costs. So, soil
needs to be tested twice a year. After this test, a farmer comes to know the
actual amount of fertilizers he should use. Realizing that there was not any
electronic, cheap, and time-saving method to detect nutrition level of soil
and hence, led to the idea for ANA.
188 Advances in Data Science and Computing Technology

14.1 INTRODUCTION

14.1.1 DETAILING ON ANA

We experience a day-by-day reality with the end goal that everything can
be controlled and worked. Therefore, yet there are relatively few huge parts
in our country where computerization has not been gotten or not been put
to an undeniable use on account of a few reasons like cost, achievability,
and so on. One such field is that of cultivating. Horticulture has been one
of the essential occupations of people since early developments and even
today manual intercessions in cultivating are unavoidable. Agro nutrition
alert (ANA) is an affordable tool for farmers which guides them to use only
required amount of chemical fertilizers, adequate irrigation and sometimes
luminosity and temperature (in case of greenhouse farming). Being small
in size, it is portable and easy to assemble. Whenever used, it simply sends
an SMS to the farmer mentioning the amount of water, fertilizers (urea,
potassium, and phosphate) required in the crop-field. In case of greenhouse
farming, according to temperature and luminosity recorded and required,
farmers can decide air exposure and light exposure inside the house.
ANA has already been fed some general crop data (all parameters for
Kharif, Rabi, and some specific commercial crops). For some specific crop
(Quitain, 1987), data can easily be fed by introducing new variables in
program (can be done by anyone with a little programming skills). ANA is
using 4G connectivity for communication, so in case of technology avail­
ability, farmers can go for GPRS/LTE communication too.

14.1.2 IMPORTANCE OF SOIL MONITORING

Plant monitoring contributes an important part in the field of agriculture

sector in our country as they can be used to grow plants under controlled
climatic conditions in order to get optimum productivity. Automating a plant
monitoring system and controlling of the climatic parameters, directly or
indirectly govern the plant growth and hence their productivity. Farmers can
even use their Nokia 1100 handset to read the messages from ANA and that
is the beauty as there is not any need of some kind of smartphones.
Soil is one of the most significant common assets. Precise investigation
of soil gives data on nature and sorts of soil (Kumar et al., 2014). The pH
value of soil is important part of soil health. The purpose of pH study is
Agro-Nutrition Alert (ANA) 189

to determine the degree of acidity or basicity of soil. The nutrients of soil

determine the acidity and basicity of a soil sample. If pH is in range of
0 to 7, then soil is acidic. If pH is 7, then soil is neutral. If pH of soil is
7 or above, then soil is basic. Yellow and red color of soil indicates the
presence of Iron-Oxide in soil. Dark black or brown color of soil indicates
the soil with high organic matter. Minerals present in the soil can also
affect the acidity of the soil. Hence, we can determine the nutrient of soil
by using color sensing. RGB are the basic colors which are arranged in
bands 321 (RGB). It denote the wavelength of electromagnetic radiations
in spectrum band. By using this RGB values we can easily determine the
nutrient composition of soil.
Samples of soil are collected and after processing, soil pH is determined
by using pH meter and NPK (Regalado and DelaCruz, 2016) values are also
determined by using chemical analysis in chemical laboratory. Soil samples
were then analyzed with ANA and RGB (Chang and Reid, 1996) color
sensor (Kimura, Togami, and Hasimoto, 2010), TCS3200, is being used for
detection of color composition of soil. The equation for pH index value for
each sample can be determined using the formulae:
pH index = R * (G/B)
From above equation value and measured soil pH value (by using
pH-meter) are compared. Similarly, to determine the soil nutrients like
Nitrogen, potassium, and Phosphorous, we determine RGB composition of
soil sample. Then, we can apply different equations to determine the NPK
content in soil.

14.2 FEASIBILITY STUDY

As we know, even an excellent idea is worthless, if it is not feasible. Feasibility

is the first and foremost study to take in account. So, let us compare ANA’s
features with manual soil testing on below bases:

14.2.1 COST

In traditional method, farmers have to visit nearby labs (within 100 km area)
and pay a sum of almost Rs. 500 for the test. This test needs to be carried
twice a year. ANA costs Rs. 2,000 (one time investment) and can help a
190 Advances in Data Science and Computing Technology

group of farmers. So, if a group of four farmers have ANA, it costs only the
price of one sample/farmer.

14.2.2 TIME CONSUMPTION

While visiting the soil testing lab, minimum two days are required to carry
out the process and generating the report. ANA requires a maximum of 15
minutes to setup and getting the digital report (via SMS).

14.2.3 ACCURACY

From district laboratory, accuracy cannot be determined as a small sample

represents acres of crop-field. However, after testing at 3–5 points in a certain
field, ANA promises 80–90% accuracy.

14.2.4 LIFETIME

We cannot compare the life of a laboratory with that of a tool. ANA has an
approx. lifetime of 7–8 years (must be handled with care).

14.3 SYSTEM DESIGN

14.3.1 HARDWARE REQUIRED

• Arduino UNO R3 microcontroller board (Chavan et al., 2018);

• LDR (4.7 kΩ);
• RGB color sensor (TCS3200);
• LM35 temperature sensor (0–500°C);
• Moisture sensor;
• GSM/GPRS module (SIM800A);
• Jumpers and connecting wires.

14.3.2 SOFTWARE REQUIRED

• Arduino IDE;
• Windows 10 (Operating system/64-bit).
Agro-Nutrition Alert (ANA) 191

14.3.3 BLOCK DIAGRAM (Figure 14.1)

FIGURE 14.1 Diagram showing function of design proposed.

14.3.4 CIRCUIT DIAGRAM (Figures 14.2–14.4)

FIGURE 14.2 Diagram showing different sensor connection to collect chemical parameters
like nitrogen, phosphorous, and potassium content (primary).
192 Advances in Data Science and Computing Technology

FIGURE 14.3 Diagram showing different sensor connection to collect atmospheric and soil
parameters like ambient light, temperature, and soil moisture (secondary).

FIGURE 14.4 Diagram showing module connection to transmit collected parameters in

form of an SMS (communication).
Agro-Nutrition Alert (ANA) 193

14.4 RESULTS

14.4.1 SERIAL MONITOR (Figure 14.5)

FIGURE 14.5 Diagram showing output in serial monitor of Arduino.

14.4.2 SMS FORMAT (Figure 14.6)

FIGURE 14.6 Diagram showing output message screenshot in recipients’ mobile phone.
194 Advances in Data Science and Computing Technology

14.5 CONCLUSION

Using the proposed solution, farmers can get their soil report within minutes.
They can test the soil at different positions and hence increase the accuracy
significantly. With its low cost and less time consumption, it makes the
process smooth and easy to carry. With the easy-to-understand SMS format
(in English + regional language), it satisfies the user up to a far better extent.
We can conclude with a hope that in near future, chemical laboratories could
be replaced through the instruments like ANA.

KEYWORDS

• agro nutrition
• agro nutrition alert (ANN)
• Arduino
• nitrogen
• phosphorous
• potassium

REFERENCES

Chang, C. Y., & Reid, J. F., (1996). RGB Calibration for color image analysis in machine vision
Image analysis in machine vision. IEEE Transactions on Image Processing, 5, 1414–1422.
Chavan, S., Badhe, A., Kharadkar, S., Ware, R., & Kamble, P., (2018). IOT based smart
agriculture and soil nutrient detection system. International Journal on Future Revolution in
Computer Science & Communication Engineering, 4(4).
Kumar, V., Vimal, B. K., Kumar, R., Kumar, R., & Kumar, M., (2014). Determination of soil
pH by using digital image processing technique. Journal of Applied and Natural Science,
6(1), 14–18.
Quitain, B. M., David, R. M. (1987). Reability of soil test kit (Philippines). Bureau of Soils,
Manila (Philippines). Philippine Journal of Crop Science, 12(1), 21.
Regalado, R. G., & DelaCruz, J. C., (2016). Soil pH and Nutrient (Nitrogen Phosphorous
and potassium) analyzer using colorimetry. Proceedings of IEEE Region 10 International
Conference – TENCON.
Yoshitsugu Kimura, Kyosuke Yamamoto, Togami, T., Hashimoto, A., Kameoka, T., &
Yoshioka, Y. (2010). Construction of the Prototype System for the Chromatic Image
Analysis Using Color Distribution Entropy, 2438–2442.
 PART V

Optimization Technique

Using MATLAB Platform

CHAPTER 15

SELECTION OF OPTIMAL MOTHER

WAVELET FOR FAULT ANALYSIS IN
INDUCTION MOTOR USING STATOR
CURRENT WAVEFORM
ARUNAVA KABIRAJ THAKUR,1 PALASH KUMAR KUNDU,2 and
ARABINDA DAS2
1
Department of Electrical Engineering, Techno Main Salt Lake, Sector V,

Bidhannagar, Kolkata–700091, West Bengal, India,

E-mail: [email protected]

2
Department of Electrical Engineering, Jadavpur University,

Kolkata–700032, West Bengal, India

ABSTRACT

Current signature analysis is used successfully for induction motor fault

detection. FFT based spectrum analysis of current signal is applied for fault
detection but it cannot be applied for detection of non-stationary signals.
Signal analysis using Wavelet transform provides time and frequency
domain information both of signals. It is widely used for signaled noising,
signal compression and features extraction. Here, mother wavelet has been
selected from different discrete wavelet transform (DWT) families using
wavelet coefficients of three phase stator current signals. The data samples
of stator current are collected from different faulty (electrical and mechanical
faults) induction motors to perform the analysis. Different mother wavelets
(db(3–10), sym(3–8), coif(1–5)) are employed for signal decomposition of
the distorted stator current signals of different faulty induction motors which
198 Advances in Data Science and Computing Technology

are called the wavelet coefficients of the signals. The decomposed signals
are assembled back using reconstruction program and square root of mean
square error (RMSE) are found for every mother wavelet. The correlation
coefficients are also calculated for each mother waveform individually from
the original signal and the coefficients of different mother wavelets. The
optimal mother wavelet are selected comparing the values of RMSE and
correlation coefficients of different mother wavelet. The optimal mother
wavelet are found among the 19 mother wavelets which has smallest RMSE
value and largest value of correlation coefficient for all faulty conditions and
for all phases.

15.1 INTRODUCTION

Fault prediction of induction motor is an important task for modern industry

to prevent the industries suffering from downtime cost and lost revenue.
The faults have been categorized according to the different components
of induction machines-stator, bearing, rotor, and other faults (Singh and
Saad, 2003). Various techniques have been applied previously for different
faults detection in induction motor (Bhavsar and Patel, 2013; Tavner, 2008;
Sangheeta, and Hemamalini, 2017). Among them, signature analysis of
stator current is very useful technique to detect faults and to localize different
abnormal conditions occurred due to different electrical and mechanical
faults in the induction motor (Benbouzid, 2000; Blödt and Rostaing, 2008;
Henao, Razik, and Capolino, 2005; Mehala, and Dahiya, 2007; Benbouzid,
Vieira, and Theys, 1999; Singh, Kumar, and Kumar, 2014; Kalaskar and
Gond, 2014; Gómez and Sobczyk, 2013). MCSA can detect the different
motor problems at an early stage to prevents it from secondary damage and
complete failure (Henao, Razik, and Capolino, 2005; Amel, Laatra, Sami,
and Nourreddine, 2013). MCSA is a less expensive technique than others
because electrical signals are cheaper, simpler, and easier to measure and
multiple sensors are not required in this technique. Fast Fourier transform
(FFT) usually converts time to frequency of signals and vice versa but it does
not provide simultaneous time and frequency information. It has limited time
duration and frequency bandwidth hence it is not efficient for representing
discontinuities. It represents a signal with a few coefficients. Short term
Fourier transform (STFT) can be applied for no stationary signal analysis
which gives time-frequency informational both but there is a problem with
STFT because for all frequencies it provides constant resolution using the
same window for entire signal analysis. To overcome the drawback of
Selection of Optimal Mother Wavelet for Fault Analysis 199

FFT and STFT the wavelet transform is used for accuracy in fault analysis
and it has gained widespread acceptance in signal processing (Yan, Gao,
and Chen, 2014). It has been shown that wavelet analysis can improve the
fault diagnosis (Ye, Wu, and Sadeghian, 2003; Poshtan and Zarei, 2007).
Discrete wavelet transform (DWT) decomposes the signal into several
wavelet coefficients (‘approximation’ and ‘detail’) and the decomposed
coefficients can be implemented by using a filter (Ponci, et al., 2007; Burrus,
Gopinath, and Guo, 1998; Bhola et al., 2013). Monique (1997) compared
between non-orthogonal and orthogonal wavelet and he has shown that
usually orthogonal wavelet have better de-noising performances than
the non-orthogonal. Orthogonal wavelet decomposition is useful for the
non-stationary signal, meaning that its variance depends on the window of
the data under consideration. Its decomposition provides information on the
variability of wave height at the various timescales with time (Daviu et al.,
2006). The improvement of de-noising performances with the increase of
data length has been shown for all thresholding schemes and transforms.
There are different orthogonal wavelets (Haar, Daubechies (db), Symlets,
Coiflets (Coif)) which are normally used for machine fault analysis. Despite
the satisfactory performance of DWT, the selection of optimal mother
wavelet is challenging task for different mechanical and electrical faults in the
machines. Researchers are applied different techniques previously to select
the optimal mother wavelet for detect of one particular fault in the machines
using vibration signals (Rafiee et al., 2009; Wang and McFadden, 1995) but
they are not selected optimal mother wavelet for analysis of more than one
type of fault using current signals. Optimal mother wavelet is selected in this
work among 19 mother wavelet functions for the thee phase current signals
of six different faulty induction motors. Haar (db1) wavelet is not used
here because the wavelet family is formed by sequence of rescaled ‘square-
shaped’ functions together and it is not continuous and hence it cannot be
used for distorted continuous signal analysis. Distorted stator currents are
decomposed using MATLAB program applying different mother wavelets
functions (db(3–10), sym(3–8), coif(1–5)) for different faulty conditions and
the decomposed signals are assembled back using reconstruction program.
The square root of mean square error (RMSE) are found from actual and
reconstructed signals and the coefficients of correlation are also calculated
using the actual signal and denoised signal of different mother wavelets
for each phase of each faulty motor. The optimal mother wavelet has been
selected based on the lowest RMSE and highest correlation coefficient value
of each phase of each faulty I.M.
200 Advances in Data Science and Computing Technology

15.2 METHOD

15.2.1 WAVELET ANALYSIS

Wavelet transform is normally used to overcome the drawback of FFT

and STFT. Wavelet transform has become one of the fast-evolving signal
processing, emerging, and mathematical tools in the recent years for many
distinct merits and the MRA led to the wavelet transform famous for its
simplicity and recursive filtering algorithm to compute the wavelet decom­
position of the signal from its finest scale approximation. It has gained
widespread acceptance for fault analysis (Aktas and Turkmenoglu, 2010) in
the machines for signal decomposition into a set of basis functions providing
variable window size, the basic functions are called wavelets. Applying
translations and scaling (stretch/compress) on the ‘mother’ wavelet ψ(t) the
basis can be constructed.

ψ t,T ( t )=
1
S
( )
ψ t-T
S (1)

where; ‘S’ and ‘t’ are the scaling and shifting parameter, respectively.
Wavelet transform is separately computed for different time-domain signal
segments at different frequencies. Wavelet coefficients are obtained, at a first
level of signal decomposition by applying a mother wavelet. The repetition
of this process depends on the scaling and translation of mother wavelet.
The mother is named due to formation the basis for various processes of
transformation. A mother wavelet can be imagined as a windowed function
which moves/shifts along the signal of time-series from the time t = 0 to
t =T. In the window the portion of the signal is multiplied by the mother
wavelet and then it is integrated over all times to get the coefficients of
wavelet. Wavelet family can represent that signal in as few coefficients as
possible is generally considered to be best suited for a particular application.
Every application would have one mother wavelet that would be most suited
to it. In wavelet theory it is tried to decompose the spectra into high and
low frequency content in some peculiar way for making some information
extraction process simpler. DWT decomposes the signal into a set of mutu­
ally orthogonal wavelet basis functions.
The functions of wavelet are dilated, translated, and scaled versions of
a common function Ψ which is known as the mother wavelet. There are
different mother wavelet functions (db), Coiflets (Coif), Symlets (Sym))
which are used for different fault detection in the machines. In DWT based
Selection of Optimal Mother Wavelet for Fault Analysis 201

decomposition the results in useful data contained in ‘approximate’ and

‘details’ parts as shown in Figure 15.1. The low frequency signal compo­
nents called ‘approximate’ and the high frequency signal components called
‘detail.’ Computing the ‘n’ level decomposition higher detail parts being
removed, thereby the overall frequency of the resulting data is reduced. The
DWT is implemented in this work to decompose a current signal into scales
with resolution of different time and frequency using a multi-resolution
signal decomposition algorithm.

FIGURE 15.1 m-th level wavelet decomposition.

15.2.2 CURRENT SIGNAL RECONSTRUCTION BASED ON

DECOMPOSITION OF WAVELET

The decomposition analysis is described in DWT and the other half of the
story is how the actual signal comes back into original form by assembling
the decomposed components of the signal without losing any information.
This name of the process is called synthesis of reconstruction. The math­
ematical manipulation of the process of reconstruction is called the inverse
discrete wavelet transform (IDWT). The process of wavelet reconstruction
consists of filtering and upsampling. Upsampling (or interpolation) is done
by zero padding between every two coefficients. The filters design for
decomposition and reconstruction is based on the well-known technique
called ‘Quadrature Mirror Filters.’ The reconstructed approximations and
202 Advances in Data Science and Computing Technology

details are the actual constituents of original signal. In fact, it is found

when the approximate coefficient vector cA1 and detail coefficient cD1 are
combined because the coefficient vectors are produced by the help of down
sampling. Before combining the approximations and details reconstruction
is necessary. The process can be extended to the multi-level component
analysis; it is found the similarity in relationships hold for all the recon­
structed signal constituents. That is, the original signal can be reassembled
by several ways.

15.2.3 STATISTICAL ANALYSIS

The signals of current and vibration from the rotating machine carry lot
of information about different fault conditions. Statistical parameters for
different fault conditions can be found by statistical analysis of the acquired
current signals. The parameters are called statistical features carry informa­
tion of time domain signals about the different fault conditions. The features
are also used to detect faults and some of the features can be used individu­
ally to distinguish between healthy and faulty components. One statistical
tool called correlation analysis, which can be used to select optimal mother
wavelet for fault analysis of the induction motor. The word ‘correlation’
denotes the degree of association between the variables. The two variables q
and r are called to be correlated if the variations of magnitude of a variable
tend to be accompanied by magnitude variations of the other variable. The
two variables are called positively correlated if r increases when q trends to
increase. The two variables are called negatively correlated if r increases as
q tends to decrease. The two variables are called uncorrelated if the values
of q are not affected by changes in the values of r. R is the matrix of correla­
tion coefficients calculated from an input matrix P as input whose rows and
columns are observations and variable respectively. R matrix is related to the
covariance matrix CV = cov (P) by:
CV ( q,r )
R( q,r ) = (2)
CV ( q,q )CV ( r,r )

Correlation technique has been used to calculate correlation coefficient

from original signal and denoised signal and to detect optimal mother
wavelet (Al-Qazzaz et al., 2007). The block diagram represents the steps
of correlation coefficients from different mother wavelet family has been
shown in Figure 15.2.
Selection of Optimal Mother Wavelet for Fault Analysis 203

FIGURE 15.2 The correlation block diagram between the noisy and de-noised signals using
mother wavelet families.

15.2.4 SELECTION OF OPTIMAL MOTHER WAVELET

Selection of optimal mother wavelet function is an essential task for wavelet

analysis for demonstration of its advantages in coefficient reconstruction,
denoising, feature extraction and component separation from the time domain
and frequency domain signal both. This transform involves correlating
between the signal being analyzed and a prototype wavelet function. Thus,
the selection of mother wavelet function can influence the fault analysis
performance of induction motor. There is a variation of qualities of different
mother wavelets criteria viz., (i) symmetry, (ii) speed of convergence to zero,
(iii) vanishing moments number, (iv) regularity, (v) existence of scaling
function, etc. Optimal mother wavelet selection is also essentials to find the
accuracy of fault analysis.
In this work optimal mother wavelet has been selected comparing the
values of correlation coefficients and root mean square error (RMSE). The
current sample of original and reconstructed signal has been used for anal­
ysis and square RMSE has been computed. The mean squared error (MSE)
measures the average of the squares of deviations or errors, i.e., the differ­
ence between the original and reconstructed signal. The RMSD or RMSE is
found taking the square root of MSE. For an unbiased estimator the RMSE
has the same units as the quantity being estimated. The RMSE is also known
as the standard deviation (SD) as it is the square root of the variance. Mean
square error and root mean square error can be defined as:
M
1
Mean square error =
M ∑[ x( p) − x( p)]
n=1
2
(3)
204 Advances in Data Science and Computing Technology

M
1
Root mean square error =
M ∑[x( p) − x( p)]
n=1
2
(4)

where; x̃  (p) is the p-th sample data of current signature for a given type of
fault and x̃  (p) is the p-th sample data wavelet filtered current signature for
the above type of fault.
The correlation coefficient (R) measures the strength of association
between the two variables. In the first step it studies the relationship between
two continuous variables and it draws a scatter plot of the two variables to
check linearity. The calculation of correlation coefficient depends on the
linear relationship. It really does not matter for correlation about the plotting
of variables in the axes. However, on the x-axis the independent (explanatory)
variable is plotted and, on the y-axis, the dependent (response) variable is
plotted conventionally. The higher strength of association between the
variables depends on how nearer the scatter points are to a straight line. It does
not matter about the units are used for measurement. Computed correlation
coefficients and computed RMSE using decomposition by wavelet transform
and reconstructed by inverse wavelet transform are compared for selecting
optimum mother wavelet. The coefficients of different mother wavelets
(Daubechies(db3-db10), Coiflets(Coif1-Coif-5), Symlets(Sym3-Sym8)) are
used for correlation analysis. Correlation coefficients are computed using
noisy and denoised signal. The mother wavelet is optimal which has minimum
RMSE value and largest correlation coefficient. The steps for optimal mother
wavelet selection have been described in flowchart as shown in Figure 15.3.

15.3 EXPERIMENTAL SET UP

As shown in Figure 15.4, laboratory test set up is made for experimental

work with inverter fed three phase induction motor (2 pole, 60 Hz, 1/2 H.P,
3,450 rpm) acts as a prime mover of a dc generator. Four incandescent bulbs
are used as loads the dc generator. The three phase currents of stator of six
induction motors with different fault conditions are used for the analysis,
viz. Bearing fault, Broken rotor bar, Misaligned rotor, Stator winding fault,
Unbalanced rotor, Single phase voltage unbalance. One three phase digital
power analyzer is used to capture three phase currents of each faulty induc­
tion motor. An essential data acquisition system is used to interface with
motor-generator system to PC through USB cable. The captured three phase
currents of each induction motors are converted to numerical data samples
(amplitude vs time) by the help of computer software and the data samples
Selection of Optimal Mother Wavelet for Fault Analysis 205

are stored into pc. The current wave of each phase of each faulty induction
motor is converted to 1,002 numerical data sample and stored as CSV file.

FIGURE 15.3 (i) Computation of correlation coefficients and RMSE using decomposition
by wavelet transform and reconstruction by inverse wavelet transform; (ii) selection of
optimum mother wavelet function.

15.3.1 DATA ACQUISITION SYSTEM

One digital three phase power analyzer (Yokogawa WT 500) is used as a

data acquisition system which captures the three phase stator currents from
induction motors with different fault conditions and it interfaces with PC
through USB cable as shown in Figure 15.5. Zero-Flux (TM) current sensors
are used in this system because it allows the precise measurement of currents
providing accurate phase-shift information. The system (WT 500) provides
and supports a wide range of voltage from 1.5 V to 1,000 V. The sampling
time is set to 5 ms for data acquisition task. The numerical values of current
206 Advances in Data Science and Computing Technology

waves are essential for the analysis. Therefore, the WT viewer software is
used in the PC to convert the current waveforms to numerical value and the
data samples are stored into.’CSV’ file format.

FIGURE 15.4 Laboratory experimental set-up.

FIGURE 15.5 The diagram of data acquisition by Yokogawa WT 500.

15.4 RESULT AND ANALYSIS

The data samples (amplitude vs. time) of three phase currents are collected
from six induction motors with different types of fault (bearing fault, broken
rotor bar, misaligned rotor, unbalanced rotor, stator winding fault, single phase
voltage unbalance) for the analysis. The wavelet coefficients are computed
using different mother wavelet functions (db(3–10), coif(1–5), sym(3–8)) from
the data samples of phase currents of each faulty induction motor by MATLAB
routine. The signals are reconstructed from different ‘approximate’ coefficients.
Root mean square error (RMSE) has been computed from original and recon­
structed signal for different mother wavelets in different faulty conditions.
Correlation coefficients are also computed for original signal and denoised
signal using MATLAB routine. The histograms (Figures 15.6–15.23) are plotted
Selection of Optimal Mother Wavelet for Fault Analysis 207

using RMSE and correlation coefficient values in different faulted conditions

for mother wavelet families. From the histograms and the tables (Tables 15.1
and 15.2), it may be concluded that ‘Symlet 5’ has minimum RMSE value and
has maximum correlation coefficients for all phases in all faulty conditions.

FIGURE 15.6 (i), (ii) and (iii) shows the correlation coefficients and RMSE between original
and reconstructed current waveform on the basis of Sym(3–8) mother wavelet functions under
fault condition as broken rotor bar for R, Y, and B phase.

FIGURE 15.7 (i), (ii), and (iii) shows the correlation coefficients and RMSE between
original and reconstructed waveform on the basis of db(3–10) mother wavelet functions
current under fault condition as broken rotor bar for R, Y, and B phase.

FIGURE 15.8 (i), (ii), and (iii) shows the correlation coefficients and RMSE between
original and reconstructed waveform on the basis of coif(1–5) mother wavelet functions
current under fault condition as broken rotor bar for R, Y, and B phase.
208 Advances in Data Science and Computing Technology

FIGURE 15.9 (i), (ii), and (iii) shows the correlation coefficients and RMSE between
original and reconstructed waveform on the basis of sym(3–8) mother wavelet functions
current under fault condition as faulted bearing for R, Y, and B phase.

FIGURE 15.10 (i), (ii), and (iii) shows the correlation coefficients and RMSE between
original and reconstructed waveform on the basis of db(3–10) mother wavelet functions
current under fault condition as faulted bearing for R, Y, and B phase.

FIGURE 15.11 (i), (ii), and (iii) shows the correlation coefficients and RMSE between
original and reconstructed waveform on the basis of coif(1–5) mother wavelet functions
current under fault condition as faulted bearing for R, Y, and B phase.
Selection of Optimal Mother Wavelet for Fault Analysis 209

FIGURE 15.12 (i), (ii), and (iii) shows the correlation coefficients and RMSE between
original and reconstructed waveform on the basis of sym(3–8) mother wavelet functions
current under fault condition as rotor misalignment for R, Y, and B phase.

FIGURE 15.13 (i), (ii), and (iii) shows the correlation coefficients and RMSE between
original and reconstructed waveform on the basis of db(3–10) mother wavelet functions
current under fault condition as rotor misalignment for R, Y, and B phase.

FIGURE 15.14 (i), (ii), and (iii) shows the correlation coefficients and RMSE between
original and reconstructed waveform on the basis of coif(1–5) mother wavelet functions
current under fault condition as rotor misalignment for R, Y, and B phase.
210 Advances in Data Science and Computing Technology

FIGURE 15.15 (i), (ii), and (iii) shows the correlation coefficients and RMSE between
original and reconstructed waveform on the basis of sym(3–8) mother wavelet functions
current under fault condition as rotor unbalance for R, Y, and B phase.

FIGURE 15.16 (i), (ii), and (iii) shows the correlation coefficients and RMSE between
original and reconstructed waveform on the basis of db(3–10) mother wavelet functions
current under fault condition as rotor unbalance for R, Y, and B phase.

FIGURE 15.17 (i), (ii), and (iii) shows the correlation coefficients and RMSE between
original and reconstructed waveform on the basis of coif(1–5) mother wavelet functions
current under fault condition as rotor unbalance for R, Y, and B phase.
Selection of Optimal Mother Wavelet for Fault Analysis 211

FIGURE 15.18 (i), (ii), and (iii) shows the correlation coefficients and RMSE between
original and reconstructed waveform on the basis of sym(3–8) mother wavelet functions
current under fault condition as stator winding fault for R, Y, and B-phase.

FIGURE 15.19 (i), (ii), and (iii) shows the correlation coefficients and RMSE between
original and reconstructed waveform on the basis of db(3–10) mother wavelet functions
current under fault condition as stator winding fault for (a) R, Y, and B-phase.

FIGURE 15.20 (i), (ii), and (iii) shows the correlation coefficients and RMSE between
original and reconstructed waveform on the basis of coif(1–5) mother wavelet functions
current under fault condition as stator winding fault for R, Y, and B phase.
212 Advances in Data Science and Computing Technology

FIGURE 15.21 (i), (ii), and (iii) shows that correlation coefficients and RMSE between
original and reconstructed waveform on the basis of sym(3–8) mother wavelet functions
current under fault condition as single phase voltage unbalance for R, Y, and B phase.

FIGURE 15.22 (i), (ii) and (iii) shows that correlation coefficients and RMSE between
original and reconstructed waveform on the basis of db(3–10) mother wavelet functions
current under fault condition as single phase voltage unbalance for R, Y, and B phase.

FIGURE 15.23 (i), (ii), and (iii) shows that correlation coefficients and RMSE between
original and reconstructed waveform on the basis of coif(1–5) mother wavelet functions
current under fault condition as single phase voltage unbalance for R, Y, and B phase.
 Selection of Optimal Mother Wavelet for Fault Analysis
TABLE 15.1 Correlation Coefficients between Original and Reconstructed Waveform Using Mother Wavelet Sym5 Under Different Faulty
Conditions for R-Y and B Phases
Mother Wavelet BRF BRB RML RTU SWF FVU
R 0.996827021 0.998806583 0.996141276 0.996788131 0.996660663 0.996883995
Y 0.996944169 0.999697144 0.996869577 0.996625068 0.996039162 0.996844161
B 0.996806649 0.999756636 0.997383615 0.997378658 0.99723101 0.997124796

213
 214
TABLE 15.2 Root Mean Square Error between Original and Reconstructed Waveform Using Mother Wavelet Sym5 Under Different Faulty
Conditions for R-Y and B Phases

Advances in Data Science and Computing Technology

Mother Wavelet BRF BRB RML RTU SWF FVU
R 0.140121385 0.04173592 0.15925463 0.140197881 0.150676488 0.140892016
Y 0.14689829 0.039426956 0.150351419 0.153585049 0.16056938 0.144927614
B 0.14163817 0.03674286 0.13884757 0.130144044 0.13832768 0.141638165
Note: BRB: Broken rotor bar; BRF: Bearing fault; RML: Misaligned rotor; RTU: Unbalanced rotor; SWF: Stator winding fault; FVU: Single
phase voltage unbalance.
Selection of Optimal Mother Wavelet for Fault Analysis 215

15.5 DISCUSSION AND CONCLUSION

Optimal mother wavelet selection is an essential task for accuracy in different

types of electrical and mechanical fault detection of induction machine using
current analysis. It is the critical task to choose the optimal mother wavelet
from different mother wavelet for induction motor fault analysis. In this
work the optimal mother wavelet has been identified for fault detection in
the induction motor analyzing the three phase stator currents. The original
distorted current signal from different faulty induction motor and the denoised
current signal using different mother wavelet have been compared and root
mean square error (RMSE) has been found. Comparing the RMSE and the
values of correlation coefficient it has been found that the Symlet5 mother
wavelet is the optimal one for detection of electrical and mechanical type
faults in the induction motor. As shown in the tables for all phases Sym5 has
minimum RMSE values and the maximum values of correlation coefficient
than other wavelet and hence it can be said that Sym5 is the optimal one.

KEYWORDS

• correlation coefficient
• discrete wavelet transform (DWT)
• mother wavelet selection
• root of mean square error (RMSE)

REFERENCES

Aktas, M., & Turkmenoglu, V., (2010). Wavelet based switching fault detection in direct torque
control induction motor drives. IET Science, Measurement, and Technology, 4(6), 303–310.
Al-Qazzaz, N. K., Ali, S. H. B., Ahmad, M. S. A., Islam, M. S., & Escudero, J., (2007). Selection
of mother wavelet function for multi-channel EEG signal analysis during a working memory
task. Sensors, 15(11), 497–507.
Amel, B., Laatra, Y., Sami, S., & Nourreddine, D., (2013). Classification and diagnosis
of broken rotor bar faults in induction motor using spectral analysis and SVM. In: 8th
International Conference and Exhibition on Ecological Vehicles and Renewable Energies
(EVER) (pp. 1–5).
Benbouzid, M. E. H., (2000). A review of induction motors signature analysis as a medium for
faults detection. IEEE Transactions on Industrial Electronics, 47(5), 983–993.
216 Advances in Data Science and Computing Technology

Benbouzid, M. E. H., Vieira, M., & Theys, C., (1999). Induction motors fault detection and
localization using stator current advanced signal processing techniques. IEEE Transactions
on Power Electronics, 14(1), 14–22.
Bhavsar, R. C., & Patel, R. A., (2013). Various techniques for condition monitoring of three
phase induction motor: A review. International Journal of Engineering Inventions, 3(4),
22–26.
Bhola, J., Yadav, R. B., Rao, K. R. M., & Yadav, H. L., (2013). Selection of optimal mother
wavelet for fault detection using discrete wavelet transform. International Journal of Advanced
Research in Electrical, Electronics and Instrumentation Engineering, 2(6), 2338–2343.
Blödt, M., & Rostaing, G., (2008). Models for bearing damage detection in induction motors using
stator current monitoring. IEEE Transactions on Industrial Electronics, 55(4), 1813–1822.
Burrus, C. S., Gopinath, R. A., & Guo, H., (1998). Introduction to Wavelets and Wavelet
Transforms. A Primer, Englewood Cliffs, NJ: Prentice-Hall.
Daviu, A. J., et al., (2006). Validation of a new method for the diagnosis of rotor bar failures
via wavelet transformation in industrial induction machines. IEEE Transactions on Industry
Applications, 42(4), 990–996.
Fargues, M. P., Barsanti, R. J., & Hippenstiel, R., (1997). Wavelet-based denoising:
Comparisons between orthogonal and non-orthogonal decompositions. Proceedings of 40th
Midwest Symposium on Circuits and Systems (pp. 929–932). Dedicated to the Memory of
Professor Mac Van Valkenburg.
Gómez, A. J. F., & Sobczyk, T. J., (2013). Motor current signature analysis apply for external
mechanical fault and cage asymmetry in induction motors. In: 9th International Symposium
on Diagnostics for Electric Machines, Power Electronics and Drives (SDEMPED) (pp.
136–141).
Henao, H., Razik, H., & Capolino, G. A., (2005). Analytical approach of the stator current
frequency harmonics computation for detection of induction machine rotor faults. IEEE
Transactions on Industry Applications, 41(3), 801–807.
Kalaskar, C. S., & Gond, V. J., (2014). motor current signature analysis to detect the fault in
induction motor. International Journal of Engineering Research and Applications, 4(6),
58–61.
Mehala, N., & Dahiya, R., (2007). An approach of condition monitoring of induction motor
using MCSA. International Journal of Systems Applications, Engineering & Development,
1(1), 13–17.
Ponci, F., et al., (2007). Diagnostic of a faulty induction motor drive via wavelet decomposition.
IEEE Transactions on Instrumentation and Measurement, 56(6), 2606–2615.
Poshtan, J., & Zarei, J., (2007). Bearing fault detection using wavelet packet transform of
induction motor stator current. Tribology International, 40(5), 763–769.
Rafiee, J., Tse, P. W., Harif, A., & Sadeghi, M. H., (2009). A novel technique for selecting
mother wavelet function using an intelligent fault diagnosis system. Expert Systems with
Applications, 36(3), 4862–4875.
Sangheeta, P., & Hemamalini, S., (2017). Dyadic wavelet transform-based acoustic signal
analysis for torque prediction of a three phase induction motor. IET Signal Processing,
11(5), 604–612.
Singh, G. K., & Sa’ad, A. S. A. K., (2003). Induction machine drive condition monitoring
and diagnostic research—A survey. Electrical Power Systems Research, 64(2), 145–158.
Singh, S., Kumar, A., & Kumar, N., (2014). Motor current signature analysis for bearing fault
detection in mechanical system. Procedia Materials Science, 6, 171–177.
Selection of Optimal Mother Wavelet for Fault Analysis 217

Tavner, P. J., (2008). Review of condition monitoring of rotating electrical machines. IET
Electric Power Applications, 2(4), 215–247.
Wang, W. J., & Mc Fadden, P. D., (1995). Application of orthogonal wavelets to early gear
damage detection. Mechanical Systems and Signal Processing, 9(5), 497–507.
Yan, R., Gao, R. X., & Chen, X., (2014). Wavelets for fault diagnosis of rotary machines: A
review with applications. Signal Processing, 96(Part A), 1–15.
Ye, Z., Wu, B., & Sadeghian, A., (2003). Current signature analysis of induction motor
mechanical faults by wavelet packet decomposition. IEEE Trans on Industrial Electronics,
50(6), 1217–1227.
CHAPTER 16

HIGH EFFICIENCY HIGH GAIN SEPIC­

BUCK BOOST CONVERTER BASED
BLDC MOTOR DRIVE FOR SOLAR PV
ARRAY FED WATER PUMPING
K. M. ASHITHA1 and D. THOMAS2
1
Department of Electrical and Electronics Engineering,
Amal Jyothi College of Engineering, Kanjirappally, Kerala–686518,
India, E-mail: [email protected]
2
Amal Jyothi College of Engineering, Kanjirappally, Kottayam,
Kerala–686518, India

ABSTRACT

Renewable energy sources like solar power may be employed for water
pumping applications where Solar irradiation availability is plenty and
where requirement of pumped water is frequent. BLDC motor drives are
preferably employed for pumping, owing to their better power delivery,
dynamic response, and efficiency, when compared to Induction motor
and other dc drives. In this chapter, studied the performance of a DC-DC
converter to reduce overall power consumption in a Solar powered BLDC
motor drive-based water pumping system. analyze and design the converter
so as to enhance the efficiency of the system. A high efficiency, derived non-
isolated SEPIC-based converter is designed for use in between the SPV and
voltage switching inverter systems. The converter mentioned in this chapter
is derived from SEPIC based buck boost converter. P&O methods have been
220 Advances in Data Science and Computing Technology

used to collect solar energy efficiently. Brushless DC motor is controlled

by switching pattern of VSI. The necessary switching arrangements for the
voltage switching inverter are made by means of electronic commutation
methods. Mathematical design of proposed system is also explained here.
The proposed system was verified by simulation using MATLAB.

16.1 INTRODUCTION

Due to increasing population and industrialization, global energy crisis as

well as crude oil prices are rising day by day. Renewable energy sources,
fuel cells and likewise energy storage systems are resources to bridge the gap
between energy demand and supply. The devaluation of electronic devices
and solar plants may attract researchers to these areas. Solar power generation
technology is ceaseless, clean, silent, and do not release any air or water
pollution to the environment (Kuthsiyat, Chandru, Dhanapriyan, Kishore,
and Vinothraj, 2017; Siva and Sampathkumar, 2015). Among the various
applications of PV source, water pumping has been gaining significance.
Nowadays, solar powered water pumping is used for industries, agricultural
sectors, and household applications.
The conventional DC motors and induction motors are commonly used
for water pumping applications. But, low efficiency, higher acoustic noise,
complexity in control, etc., have reduced their advantages. In last decades,
the BLDC motors were used as an alternative to conventional motors. Later,
the use of BLDC motor catered to the large scale industrial and agricultural
applications, due to its unique features like better efficiency, lesser noise,
light weight, good dynamic response, longer life, etc. (Kumar and Singh,
2017; Singh and Kumar, 2015). As a rule, the BLDC motors are feed by
voltage switching inverters (VSI). In this chapter, the motor’s speed is
controlled by the variations in the inverter’s output voltage so no additional
controlling circuits are needed here.
There are many options for draining maximum energy from solar plants
which are explained by Eltawil and Zhengming Zhao (2013); Bendib,
Belmili, and Krim (2015). The important output voltage of the solar power
sources, allude to the importance of DC-DC converters. Such converters are
the intermediaries of the solar panel and the inverter. Traditional buck-boost
converters have some disadvantages like pulsating output and input current,
inverting output voltage, etc. In-order to negotiate these annoyances, the
high gain converters have come into force (Zhang, Zhang, See, and Zhang,
2018; Banaei and Sani, 2018). But power losses of high gain converters are
High Efficiency High Gain SEPIC-Buck Boost Converter 221

much more flagrant subject. Some converters are capable to providing higher
and lower output voltage levels, but it may be inverting or negative. “Voltage
gain of the fly back converter can be adjusted by its turn’s ratio” (Hwu and
Yau, 2012). But leakage inductance leads to power losses and reduces the
efficiency. “Another peculiarity of dc-dc converter is reduced voltage stress
across switches. In the absence of input side inductors this may become
large” (Miao and Gao, 2019).
Solar powered BLDC motor drive for water pumping system is presented
in this chapter. MATLAB implementation of system model with electronic
commutation is also established. This chapter is divided in to five sections.
Introduction is in Section 16.1. System configuration and system design
are discussed in Sections 16.2 and 16.3. Section 16.4 deals with the control
parameters of the system. Further results and findings are included in Section
16.5. Finally, we conclude the chapter in Section 16.6.

16.2 SYSTEM OUTLINE

Figure 16.1 shows schematic diagram of the proposed system. High effi­
ciency SEPIC based buck boost converter is an intermediate of SPV panel
and VSI.

FIGURE 16.1 Schematic diagram of entire system.

Switching pulses of High gain converter is generated by P&O MPPT

method, P&O tracks the most power from PV panel. Output of high gain
222 Advances in Data Science and Computing Technology

converter is fed into voltage switching inverter and also the BLDC motor is
driven by controlling the switching pattern of voltage switching inverter. The
switching pulses for VSI is generated by Hall Effect signal position sensors.

16.3 SYSTEM DESIGN

The system consists of PV panel, high efficiency SEPIC based buck boost
converter, voltage switching inverter, BLDC motor and centrifugal pump. A
2.2 kW centrifugal pump and 2.5 kW PV array were selected for the design
and simulation.

16.3.1 SELECTION OF PV PANEL

A 2.5 kW PV array is selected, to feed sufficient power for the 2.2 kW pump.
The output of the PV array is fed into the high gain converter. Details of PV
panel per string is given in Table 16.1. Three such panels in series make up a
string. Four such strings are connected in parallel to complete the SPV array.

TABLE 16.1 Parameters of PV System

PV Panel
Parameters Value
Maximum power (PMP) 215 W
Maximum voltage (VMP) 29 V
Maximum current (IMP) 7.35 A
Number of cells (Ns) 60
PV Array System
Maximum power (PMP) 2,557 W
Maximum voltage (VMP) 87 V
Maximum current (IMP) 29.4 A

16.3.2 SELECTION OF DC-DC CONVERTOR

The high gain, derived SEPIC based converter utilized in between solar
array and also the three-phase voltage switching inverter is shown in Figure
16.2. The converter consists of two diodes D1, D2, inductors L1, L2, L3 and
capacitors C1, C2, C01, C02. Figures 16.3(a) and (b) show the 2 modes of
High Efficiency High Gain SEPIC-Buck Boost Converter 223

operation of the converter. Mode 1 is with the switch in turned ON position

and Mode 2 is with the switch in turned OFF position.

FIGURE 16.2 Derived SEPIC-based buck boost converter.

Mode1: Switch S is turned ON, inductor L1 charges through input source

V. Inductor L2 capacitor C1, and L3 charges through C01. Here capacitor C2
is also charging. All diodes D1, D2 are reverse biased in this mode, therefore,
output capacitor C02 feeds the load. Figure 16.3(a) shows mode 1 operation
of proposed converter.
The volt-second balance equation applied to the inductors L1, L2 and L3:
VL1 = V (1)
VL2 = −VC1 (2)
VL3 =
−VC1 −VC 2 +V01 (3)
Mode 2: Switch S turned OFF, diodes D1, D2 are forward biased. Inductor
L1 discharges through C1. Capacitors C01 and C02 are charging. Related
equations are given below:
VL1 =V −VC1 −VL2 (4)
VL2 =
−VC 2 +V C 02 (5)
VL3 = −VC 2 (6)
1  DTS TS
 ∫ Vdt + ∫ (V −VC1 −VL2 ) dt  =
0 (7)
TS  0 DT S 
1  DTS TS
 ∫ −VC1dt + ∫ ( −VC 2 +VC 02 ) dt  =
0 (8)
TS  0 DTS 
224 Advances in Data Science and Computing Technology

FIGURE 16.3 (a) Mode 1 operation of proposed converter; (b) mode 2 operation of proposed
converter.
High Efficiency High Gain SEPIC-Buck Boost Converter 225

16.3.2.1 VOLTAGE TRANSFER RATIO OF THE PROPOSED

CONVERTER
DTS
1  TS 
TS  ∫ (VC1 −VC 2 +VC 01 ) dt + ∫
DTS
VC 2 dt  =

0 (9)
 0 
Substitute Eqn. (5) in Eqn. (7). We get:
V 1− D
= (10)
VC 01 D
Substitute Eqn. (10) in Eqn. (9). We have:
D
VC 2 = V (11)
1− D
Substitute Eqn. (11) in Eqn. (8):
VC 02 = 2VC 01 (12)
2D
VC 02 = V (13)
1− D
Voltage transfer ratio M is obtained:
V0 2D
M
= = (14)
V 1− D

16.3.3 CALCULATION OF VOLTAGE AND CURRENT RIPPLES OF

CONVERTOR

The amount of current changes during switching cycle is known as current

ripple through inductor, which is calculated by applying volt-second balance
equation:
di
V =L (15)
dt
Current ripple through the inductor L1, L2, L3
VD
∆I L123 = (16)
f S L1.2.3
The switching pulse frequency for the converter is taken to be sufficiently
large at about 20,000 Hz. The standardized value for current ripple is taken to
be 30% at input and output. For the designed system, the converter receives
about 87 V at its input and delivers 240 V at output, giving a Duty ratio
226 Advances in Data Science and Computing Technology

of 0.5. Voltage ripple across the capacitor is mainly due to current flowing
through the capacitor’s equivalent series resistor, and charging-discharging
of capacitor. Voltage ripple is assumed to be present at standardized value of
0.3%.
Voltage ripple through capacitor C1:
2DI
∆VC = 0 (17)
f S C1

Here I0 is the converter output current.

Voltage ripple across C2, C01, C02:

DI
∆VC = 0 (18)
f S C2.01.02

Table 16.2. shows the value of components in the converter.

TABLE 16.2 Parameters of Converter

Parameters Value
L1 3.7 × 10–4 H
L2, L3 8.1 × 10–4 H
C1 17 × 10–5 F
C2, C01, C02 8.6 × 10–5 F
fs 20,000 Hz

16.3.4 SELECTION OF DC LINK CAPACITANCE OF VSI

In order to select DC link capacitor, we determine the lowest and highest

value of VSI output voltage frequency. Let wn be the highest frequency and
wm be the lowest frequency (rad/sec):
NS × P 28000× 4
π f 2π
wn 2=
= = 2π = 586rad/sec (19)
120 120
Here f is the frequency of output voltage of VSI, NS is the maximum rated
speed of motor. “The 6th harmonic component of motor voltage appears on
DC link” (Kumar and Singh, 2017), whose corresponding ripple voltage is
ΔVd. The ripple amount is assumed to be 8% of VSI input voltage. The DC
link capacitance at this rotor speed is given as:
Id
=C = 148µ F (20)
6wn ∆Vd
High Efficiency High Gain SEPIC-Buck Boost Converter 227

The minimum speed of the motor is Nl, then wm will be:

Nl × P
π f 2π
wm 2=
= = 230rad/sec (21)
120
Corresponding value for capacitance is:
Id
=C = 377 µ F (22)
6wm ∆Vd
From the above values of capacitances, we choose the higher value of
capacitance, approximately equal to 470 μF.

16.3.5 SELECTION OF WATER PUMP

A 3 hp (approx. 2.2 kW) water pump is chosen as the system load. Its output
power is given as:
P = K w N m3 (23)

“Kw is the water pump constant (Watt/(rad/second)3) and Nm is the rotor

mechanical speed (rad/sec)” (Kumar and Singh, 2017). Mechanical torque of
the BLDC motor is 7.5 Nm at the rated speed of 2,800 rpm.
Tl
K=
w = 8.7 ×10−5 watt(rad / sed )3 (24)
N m2
We select a suitable water pump for the system with above data.

16.3.6 CONTROL OF PROPOSED SYSTEM

The switching pattern of derived converter is controlled by P&O MPPT

method. Gate pulses of VSI is produced by electronic commutation method.

16.3.7 P&O MPPT ALGORITHM

MPPT method is employed for tracking and acquiring maximum power

from the PV panel. Many algorithms to extract maximum power like perturb
and observe (P&O), Incremental Conductance, Parasitic capacitance and
Constant Voltage Methods are commonly available. In this chapter, P&O
MPPT method is chosen for the proposed system. Oscillations around the
peak power are avoided and perturbation size is 0.001.
228 Advances in Data Science and Computing Technology

16.4 ELECTRONIC COMMUTATION

BLDC motor requires three hall sensors to detect the rotor position. Hall
sensor signal values and VSI switching pattern are given in Table 16.3.

TABLE 16.3 Hall Signals and Switching Position

Hall Signals Switching Pattern
101 S1; S4
001 S1; S6
011 S3; S6
010 S3; S2
110 S5; S2
100 S5; S4

16.5 RESULT AND DISCUSSION

Figure 16.3 shows the MATLAB model of the proposed system. P&O algo­
rithm generated as a sub system is shown in Figure 16.3(a). In P&O method,
instantaneous power of PV array is compared to the previous sample and
duty for the converter is generated. Gate signals for VSI is generated as
another sub system which is shown in Figure 16.3(b). Electronic commu­
tation method is used to produce these gate signals. We measure the hall
signals generated by the motor and determine the rotor positions or rotation
by controlling the switching pattern of VSI. Motor with rated speed of 2,800
rpm and output power of 2.2 kW is simulated in this model.

16.5.1 PERFORMANCE OF THE PROPOSED SYSTEM IN

STANDARD IRRADIATION LEVEL 1,000 W⁄M2

Performance of solar PV system, high gain converter, and brushless dc motor

shown in Figure 16.3. Explanations of this are covered in further subsections.
In standard irradiation the motor reaches its maximum speed.

16.5.2 ANALYSIS OF SOLAR PV ARRAY

In steady state condition, PV array attains its maximum output power.

Maximum voltage output at irradiation 1,000 W/m2 is 87 V and maximum
High Efficiency High Gain SEPIC-Buck Boost Converter 229

current will be 29.4 A. Also, the out-put power of PV array in standard

condition is 2.5 kW which is shown in Figure 16.3(a).

16.5.3 ANALYSIS OF DC-DC CONVERTER

High gain SEPIC based buck boost converter operates in continuous conduc­
tion mode. Current through inductor L1 is continuous in nature which is
shown in Figure 16.3(b). The output power of the conductor varies with
respect to the change in irradiation level. However, converter gives maximum
output power of 2.2 kW at 1,000 W/m2 solar irradiation.

16.5.4 ANALYSIS OF BRUSHLESS DC MOTOR

Figure 16.4(c) shows the steady state performance of BLDC motor at 1,000
W/m2. Along with the DC link voltage, back EMF, and stature current
increases from zero to a steady state value it indicates, the soft starting of
motor. Electromagnetic torque produced by the BLDC motor is same as that
of load torque in steady state condition which is nearly 7.5 Nm. Also, the rotor
speed reaches its rated value in standard conditions which is approximately
2,800 rpm.

16.5.5 PERFORMANCE OF SYSTEM IN DYNAMIC STATE

Dynamic state performance of solar PV system, high gain converter, and

brushlesss dc motor are shown in Figure 16.5. Explanations of these are
covered in further subsections; the system designed irradiation levels and
time duration are listed in Table 16.4.

TABLE 16.4 Time Duration and Irradiation Level

Time in Seconds Solar Irradiation in W/m2
0–1 200
1–1.5 400
1.5–2 600
2–2.5 1,000
2.5–3 800
3–4 600
230 Advances in Data Science and Computing Technology

16.5.6 ANALYSIS OF SOLAR PV ARRAY

PV array is analyzed in different irradiation levels. The output power of

PV is nearly 500 W at 200 with an output voltage of 100 V. Output of PV
array changes with respect to change in irradiation level which is shown in
Figure 16.5(a).

FIGURE 16.4 (Continued)

High Efficiency High Gain SEPIC-Buck Boost Converter 231

FIGURE 16.4 Simulation diagram: (a) P&O MPPT; (b) electronic commutation method.

16.5.7 ANALYSIS OF DC-DC CONVERTER

High gain SEPIC based buck boost converter operates in continuous conduc­
tion mode at all different irradiation level. Current through inductor L1 is
continuous in nature which is shown in Figure 16.5(b). The output power
of the conductor varies with respect to the change in irradiation level. The
converter gives 120 V output at 200 W/m2 solar irradiation level. As we can
see from the figures the converter performs well dynamically to adjust to the
varying irradiation levels.
232 Advances in Data Science and Computing Technology

FIGURE 16.5 Steady state analysis at 1,000 W/m2: (a) PV outputs; (b) converter output;
(c) motor output.
High Efficiency High Gain SEPIC-Buck Boost Converter 233

FIGURE 16.6 Dynamic state analysis: (a) PV outputs; (b) motor output; (c) converter output.
234 Advances in Data Science and Computing Technology

16.5.8 ANALYSIS OF BRUSHLESS DC MOTOR

The dynamic performance of BLDC motor drive is shown in Figure

16.6(c). In order to analyze dynamic performances, the system is simulated
in different time duration. All the motor variables like rotor speed, back
emf, electromagnetic torque and load torque are dependent upon irradia­
tion level. The minimum running speed of motor is 1,100 rpm. Parameters
like, motor torque, power output, and efficiency are dependent upon the
irradiation level. 200 W/m2 is the minimum value of solar irradiation when
the pump is operable. At the time of minimum irradiation level, the motor
gives 1,100 rpm speed which is capable to drive the load. The efficiency
of the motor reaches 93% at 1,000 W/m2 and 88% at 200 W/m2 solar irra­
diation levels. The efficiency and output power of the motor decreases
corresponding to decrease in irradiation level which is clear from Table
16.5. The graphical representation of variation of SPV array output power,
derived SEPIC based Converter output power, and efficiency of the BLDC
motor is shown in Figure 16.7. Efficiency of motor is almost constant at
all irradiation levels. As the irradiation level changes the power output is
also changing, but the motor works at almost constant efficiency.

TABLE 16.5 Motor Performances Under Different Operating Condition

Irradiation Motor Output Motor Speed Pin of Motor ή (%)
W/m2 Power (W) (rpm) in (W)
200 460 1,000 520 88
400 900 1,600 1,000 90
600 1,350 2,000 1,480 91
800 1,850 2,400 2,000 92.5
1,000 2,230 3,800 2,400 93
800 1,850 2,400 2,000 92.5
600 1,350 2,000 1,480 91

16.6 CONCLUSION

High gain derived DC-DC converter-based solar powered water pumping

system is designed, modeled, and simulated in MATLAB software. Also, the
system was studied in steady state and dynamic conditions. High gain derived
SEPIC based buck boost converter is an intermediate of PV panel and VSI.
Reduced power losses, ease of implementation, high efficiency are the main
advantages of this converter. The performance of BLDC motor was simulated
High Efficiency High Gain SEPIC-Buck Boost Converter 235

at different irradiation levels. From the studies it is clear that the BLDC motor
is suitable for all environmental conditions. The efficiency, speed, and torque
outputs of the motor is dependent upon the irradiation level.

FIGURE 16.7 Graphical analysis of motor performance in different irradiation level.

KEYWORDS

• BLDC motor
• converter
• MATLAB
• SEPIC

REFERENCES

Aleena, P. K., Siny, P., & Babu, P., (2017). “Transformer less buck-boost converter with
positive output voltage and feedback.” International Journal of Engineering Research,
6(06). Available: 10.17577/ijertv6is060337.
Banaei, M., & Bonab, H., (2017). “A novel structure for single-switch nonisolated transformer
less buck-boost DC–DC Converter.” IEEE Transactions on Industrial Electronics, 64(1),
198–205. Available: 10.1109/tie.2016.2608321.
236 Advances in Data Science and Computing Technology

Banaei, M., & Sani, S., (2018). “Analysis and Implementation of a new SEPIC-based single-
switch buck-boost DC–DC converter with continuous input current.” IEEE Transactions on
Power Electronics, 33(12), 10317–10325. Available: 10.1109/tpel.2018.2799876.
Bendib, B., Belmili, H., & Krim, F., (2015). “A survey of the most used MPPT methods:
Conventional and advanced algorithms applied for photovoltaic systems.” Renewable and
Sustainable Energy Reviews, 45, 637–648.
Gomes De, A. B., Pacheco, C. B. E., Bitencourt, N. C., & Agostini, Jr. E., (2019). “High­
voltage-gain integrated boost-SEPIC DC-DC converter for renewable energy applications.”
Eletrônica de Potência, 24(3), 336–344. Available: 10.18618/rep.2019.3.0025.
Hwu, K. I., & Yau, Y. T., (2012). “High step-up converter based on charge pump and boost
converter.” IEEE Transactions on Power Electronics, 27(5), 2484–2494.
Kumar, R., & Singh, B., (2017). “Solar PV powered BLDC motor drive for water pumping
using Cuk converter.” IET Electric Power Applications, 11(2), 222–232. Available: 10.1049/
iet-epa.2016.0328.
Kuthsiyat, J. S., Chandru, K., Dhanapriyan, B., Kishore, K. R., & Vinothraj, G., (2017). “SEPIC
converter based water driven pumping system by using BLDC motor.” Bonfring International
Journal of Power Systems and Integrated Circuits, 7(1), 07–12. Available: 10.9756/bijpsic.8317.
Liao, H., Liang, T., Yang, L., & Chen, J., (2012). “Non-inverting buck-boost converter with
interleaved technique for fuel-cell system.” IET Power Electronics, 5(8), 1379. Available:
10.1049/iet-pel.2011.0102.
Miao, S., & Gao, J., (2019).“A family of inverting buck-boost converters with extended conversion
ratios.” IEEE Access, 7, 130197–130205. Available: 10.1109/access.2019.2940235.
Mohamed, A. E., & Zhengming, Z., (2013). “MPPT Techniques for Photovoltaic Applications.”
Renewable and Sustainable Energy Reviews, 25, 793–813.
Roberto, F. C., Walbermark, M. D. S., & Denizar, C. M., (2012). “Influence of power
converters on PV maximum power point tracking efficiency.” In: 10th IEEE/IAS Inter. Conf.
on Industry Applications (INDUSCON) (pp. 1–8).
Sarikhani, A., Allahverdinejad, B., Hamzeh, M., & Afjei, E., (2019). “A continuous input and
output current quadratic buck-boost converter with positive output voltage for photovoltaic
applications.” Solar Energy, 188, 19–27. Available: 10.1016/j.solener.2019.05.025.
Selva, K. R., Vignesh, C. J., Gayathri, D. V. P., & Naveena, P., (2016). “Design and comparison
of quadratic boost converter with boost converter.” International Journal of Engineering
Research, 5(01). Available: 10.17577/ijertv5is010650.
Singh, B., & Kumar, R., (2016). “Solar photovoltaic array fed water pump driven by brushless
DC motor using Landsman converter.” IET Renewable Power Generation, 10(4), 474–484.
Available: 10.1049/iet-rpg.2015.0295.
Siva, T., & Sampathkumar, A., (2015). “Sensor less control of brushless DC motor using
modified back-EMF detection.” Indian Journal of Science and Technology, 8(32). Available:
10.17485/ijst/2015/v8i32/87320.
Thrishna Jayaraj, & Justin Sunil Dhas, G. (2020). [Online]. Available: https://www.
researchgate.net/publication/333352436_KY_Based_C_DC_Converter_for_Standalone_
Photovoltaic_Water_Pumping_System_Employing_Four_Switch_BLDC_Drive (accessed
on 08 December 2021).
Zhang, N., Zhang, G., See, K., & Zhang, B., (2018). “A single-switch quadratic buck-boost
converter with continuous input port current and continuous output port current.” IEEE Trans­
actions on Power Electronics, 33(5), 4157–4166. Available: 10.1109/tpel.2017.2717462.
CHAPTER 17

A CLASSY FUZZY MPPT CONTROLLER

FOR STANDALONE PV SYSTEM
RAHUL KUMAR SINGH,1 RAHUL KUMAR PRASAD,1
RAHUL SINGH,1 AAYUSH ASHISH,1 DEBAPARNA SENGUPTA,1
RISHIRAJ SARKER,2 and MONALISA DATTA1
1
Techno International New Town, Kolkata–700156, West Bengal, India
2
Jadavpur University, Kolkata–700032, West Bengal, India,
E-mail: [email protected]

ABSTRACT

Maximum power point tracking (MPPT) is a powerful technique to extract

the maximum amount of power from the PV panel under different ambient
conditions. Several types of MPPT techniques have been invented and
applied by researchers in course of time to improve their performance
indices. This chapter presents a classy fuzzy logic based MPPT controller,
which is highly efficient in tracking the maximum power point within a
stipulated time period at a high accuracy. The proposed intelligent controller
is smart enough to respond very quickly to the abrupt environmental changes
and precisely tracks the maximum power point.

17.1 INTRODUCTION

Conventional energy crisis is a burning issue in today’s world. Day-by-day,

due to the upliftment of standard of living, energy demands are increasing
rapidly (Seddik et al., 2010), but the conventional energy resources are
238 Advances in Data Science and Computing Technology

diminishing in the reverse order. This has created an unresolved problem.

The demand is greater than supply. To eliminate this imbalance in supply and
demand, renewable energy resources are playing a vital role as an alternative
of fossil fuel.
In between the different types of renewable energy sources, the most
commonly used resources are solar and wind energies. Renewable energy
resources are easily available in nature in ample amount. The additional
advantages associated with the newly developed renewable energy stations
are they are eco-friendly. Thus, the various socio-economic problems like
global warming, greenhouse gas emission, pollution, health issues, etc.,
are being resolved with the promotion of renewable energy-based power
stations. Solar energy (photo voltaic energy) is the kind of nonconventional
energy, which is plentiful in nature. It has the advantage that it is being
directly converted into DC type electrical energy by the PV panel and thus
doesn’t require additional mechanism to convert the energy into DC form,
before storing it into a battery bank for future use. The main concern related
to PV based energy station is the variation in environmental condition such
as irradiation and temperature, which directly effects the power supplied by
solar PV module. If any measure for maximum Power Point tracking (MPPT)
is not initiated within the PV integrated system, the load power delivered
will not always be maximum. It will vary with the irradiance and ambient
temperature, whenever there will be deviations from the standard test condi­
tions (STC) and finally maximum power extraction will be disrupted. To
eliminate this problem, different types of MPPT controllers are used till now
(Salas et al., 2006; Esram and Chapman, 2007; Bhatnagar and Nema, 2013;
Zegaoui et al., 2011; Mirbagheri, Mekhilef, and Mirhassani, 2013; Rekioua,
Achour, and Rekioua, 2013; Oulad-Abbou et al., 2018). But these control­
lers have complicated configurations or mechanisms. Moreover, finely tuned
tracking mechanism cannot be materialized with these types of controllers.
Fuzzy logic controllers (FLCs) are robust, simple, and most importantly, it
does not require any detailed knowledge about the huge complicated system
model to be controlled before designing this type of controller. They are
capable of precisely controlling the control variable subjected to variation
in input parameters. Henceforth, Fuzzy MPPT controller is an ideal choice
for tracking the maximum power point. In 2014 (Bendib et al., 2014),
fuzzy MPPT controller integrated PV system is presented with DC-DC
buck converter. In this proposed work, a Fuzzy MPPT controller is used for
tracking the Maximum Power Point of PV module, where the system has a
DC-DC boost converter. In Villalva, Gazoli, and Ruppert (2009); Salameh,
A Classy Fuzzy MPPT Controller for Standalone PV System 239

Casacca, and Lynch (1992); Nafeh (2011); Coelho, Concer, and Martins
(20100; Hua and Shen (1997), the different components of the system like
PV array, Battery characteristics, DC-DC converters are discussed, study on
which are necessary to acquire knowledge about the system for execution of
the proposed work.

17.2 SYSTEM CONFIGURATION

The entire system configuration with the Fuzzy MPPT controller is presented
by Figure 17.1. The PV module is cascaded with a DC-DC boost converter
for charging an energy storage unit (ESU) to be used for charging electric
vehicles (EVs). After the boost converter, comes the DC link, which relates
to the ESU for charging it. The fuzzy MPPT controller is the sub system
designed for controlling the action of the boost converter by controlling
its duty cycle. Error and derivative of error are the two input signals fed
to the fuzzy MPPT controller. They are generated utilizing the voltage and
current signals coming out from the PV panel. The voltage and current
outputs are constantly varying with corresponding variations in irradiance
and temperature. Accordingly, the inputs to the fuzzy MPPT controller
are also varying. The controller has been designed in such a way that it
can generate appropriate control signal according to the variations in the
inputs. The control signal produced by the controller is fed into the gate
of the IGBT used inside the DC-DC boost converter. Thus, converter duty
cycle is being controlled very smoothly and precisely for getting desired
voltage and current signals at the boost converter output, which corresponds
to maximum power delivered to the ESU. The objective of this project is to
charge the ESU in the most efficient manner so that it can be used efficiently
for EV charging purpose. ESU will be installed in electric vehicle charging
stations (EVCSs). This is the preliminary work on a part of a huge system
to be designed further. The future scope of this work will be discussed in the
future scope section.

17.3 MATHEMATICAL MODEL OF PHOTOVOLTAIC MODULE

Figure 17.2(a) presents the equivalent circuit for solar (PV) cell. It consists
of a current source connected in parallel with a diode. RS and RSh are the
series and parallel resistances connected with the circuit. RSh is infinite and
RS is zero for an ideal PV cell.
240 Advances in Data Science and Computing Technology

FIGURE 17.1 Block diagram of the stand-alone PV system.

In actual practice, many PV cells are assembled together in series fashion

to form larger unit called PV module. A cluster of such modules are coupled
together in series/parallel to form PV arrays. These PV arrays are the back­
bone of the PV generation plant. Figure 17.2(b) presents the equivalent
circuit of PV array.

FIGURE 17.2 (a) Equivalent circuit of a PV cell; and (b) equivalent circuit of a PV array.

The equation relating voltage and current output of a PV cell is depicted

by:
  N   N 
   V + I( S )R s    (V + I  S  R s
 Np   
=I N p I ph
 
− N p Io  exp     −1 −  Np  (1)
 
  NS .α.VT    NS 
      R sh
     Np 
A Classy Fuzzy MPPT Controller for Standalone PV System 241

where; Iph is the photo current; Io is the diode’s reverse saturation current; α is
the diode ideality factor; V is the PV array output voltage of, I is the PV array
output current; Rs is the PV cell’s series resistance; Rsh is the PV cell’s shunt
resistance; NP is the parallel path counts of the PV array; NS is the PV cell
numbers connected in series in each parallel path; VT is the thermal voltage
of the PV cell.
K .T
VT = (2)
q
where; K is the Boltzmann constant (1.3805 × 10–23 J/K); T is the operational
temperature (in Kelvin); q is the electron charge (1.6 × 10–19 C).
Photo current is given by:

I ph= ( Iph −STC + Ki ( T − TSTC ) ) GG (3)

STC

where; Iph-STC is the short circuit photo current generated at STC, TSTC, and
GSTC imply the temperature and irradiance at STC. Ki is the short circuit
current coefficient of PV cell; T is the operating temperature in Kelvin; G
is the irradiance (W/m2); Ki is the short circuit current coefficient of the cell
usually provided by manufacturer.
The saturation current of a cell varies with cell temperature, which is
presented by:

Io =
( Iph −STC + Ki ( T − TSTC ) )
  (V + KV ( T − TSTC ) )   (4)
exp  OC _ STC  −1
  .á .V T  
   

where; VOC_STC is the cell open circuit voltage at STC; Kv is the open circuit
voltage coefficient.
In this proposed work, the standard test condition parameter values, TSTC,
and GSTC, are taken as 25°C and 1,000 W/m2 respectively. A PV array of
47 parallel strings with each string consisting of 15 series connected PV
modules are taken. Number of cells per module is chosen as 60. PV module
is used for study using Simulink MATLAB. The P-V and I-V characteristics
of the PV panel at STC are detected after setting the required parameters as
per the values mentioned in Table 17.1.
From the P-V characteristics, it is clearly seen that there is a specific
voltage of the PV array, where the maximum power is achieved (Figure 17.3).
That voltage is the Vmp. From the I-V curve, the corresponding current value
242 Advances in Data Science and Computing Technology

Imp can be noted down. As the PV array output voltage and current depends
on the irradiance and temperature directly, henceforth, with variations in
irradiance or temperature, the maximum power point varies. Figures 17.4(a)
and (b) present the I-V, P-V characteristics of the PV array for different
irradiances at TSTC (25°C) and different temperatures at GSTC (1,000 W/m2).
From the graphs of Figures 17.4(a) and (b), it can be observed that the current
and power changes abruptly keeping the output voltage V almost constant
with variations in irradiance. On the contrary, current value remains constant
with a great variation in voltage in case of temperature change keeping the
irradiance fixed.

TABLE 17.1 Electrical Characteristics Data of PV Module

Electrical Characteristics Values
Short-circuit current (Isc) 7.84 A
Current value maximum power point (Imp) 7.35 A
Voltage value at maximum power point (Vmp) 29 V
Open-circuit voltage (Voc) 36.3 V
Shunt Resistance (Rsh) 313.3991 ohms
Series Resistance (Rs) 0.39383 ohm
STC condition’s Maximum power (Pmax) 213.15 W
Isc’s temperature coefficient 0.102%/°C
Voc’s temperature coefficient – 0.36099%/°C
Photo current (Iph) 7.8649 A
Saturation current of diode (I0) 2.9259*exp–10
ideality factor of diode 0.98117

FIGURE 17.3 I–V and P–V characteristics of a PV module under STC.

A Classy Fuzzy MPPT Controller for Standalone PV System 243

FIGURE 17.4(a) I–V and P–V characteristics of a PV module for different irradiances at
TSTC (25°C).

FIGURE 17.4(b) I–V and P–V characteristics of a PV module for different temperatures at
GSTC (1,000 W/m2).

17.4 DC-DC BOOST CONVERTER

DC-DC converters are generally used for boosting up the voltage level
for high voltage applications. It is mainly used in PV integrated system
for proper regulation of the PV array output voltage. DC-DC boost
converters are used to step up the output voltage of the PV array, if the
load voltage requirement is higher than the voltage produced by the PV
panel. It is situated in between the PV array and the load (here the energy
storage unit (ESU)). The main purpose for installing the boost converter
is to facilitate the provision of MPPT by varying the duty cycle of the
boost converter. DC-DC boost converter’s circuit diagram is presented in
Figure 17.5.
244 Advances in Data Science and Computing Technology

FIGURE 17.5 Equivalent circuit of DC-DC boost converter.

The main components used in a DC-DC boost converter are: an inductor

(L), a capacitor (C), a control switch, i.e., IGBT (S), a diode (D) and a load
resistor (RL), This type of configuration facilitates with the higher output
voltage compared to the input voltage s (VS) fed into the input terminals
of the converter. This output voltage rest on the duty cycle of the IGBT.
Subsequently, by varying the duty cycle (or ON time) of the control switch,
the output voltage of the converter can be changed.
Parameters calculated for boost converter are:
V_(o)/V_i = 1/(1 – D) (5)

D= T_ON/T (6)

T= 1/f_sw (7)

where; D denotes the duty cycle. Vi(or VS) and V0 are the input and output
voltages of the converter. TON is the on time and T is the total time period of
the IGBT. fsw is the switching frequency (Table 17.2).

TABLE 17.2 Operating Values of Boost Converter

Electrical Characteristics Values
Time period (1/30×103) sec
Inductance 1.559e–4 H
Capacitance 2.767e–4
IGBT internal resistance 1e–3 ohms
IGBT snubber resistance 1e5 ohms
IGBT snubber capacitance infinite
Diode resistance 0.001 ohms
Diode forward voltage 0.8 V
A Classy Fuzzy MPPT Controller for Standalone PV System 245

17.5 FUZZY BASED MPPT

Among the various techniques articulated for tracking the maximum power
point of the PV integrated system, the modest and prevalent technique is
the perturb and observe (P&O) technique. This method facilitates good
results under steady state condition where there are no such considerable
variations in irradiation and temperature. But if there are huge variations in
these parameters at a frequent and faster rate, this method fails to track the
maximum power point precisely and quickly.
Fuzzy control is a newly developed heuristic method that facilitates the
construction of nonlinear controllers from vague information collected by
the expert. Moreover, the controller design does not require to know the
system model to be controlled accurately. The knowledge and expertise of
the designer play vital role in the successful implementation of these types of
controllers. FLC is more robust than conventional controller (Figure 17.6).
FLC can be classified three steps:
• fuzzification;
• inference engine with proper rule base; and
• defuzzification.

FIGURE 17.6 Block diagram of fuzzy MPPT controller.

17.5.1 FUZZIFICATION

“Fuzzification” is the technique of converting crisp inputs into fuzzy or

linguistic inputs. For this proposed Fuzzy MPPT controller, two inputs are
246 Advances in Data Science and Computing Technology

taken namely error (E) and change in error (CE). The membership functions
are assigned to both inputs with the help of linguistic variables using five
fuzzy subsets called negative big (NB), negative small (NS), zero (ZE), posi­
tive small (PS) and positive big (PB). The voltage (V) and current (I) outputs
from the PV array is constantly monitored by the Fuzzy MPPT controller and
instantaneous power is calculated. The governing equations of the inputs E
and CE to the controller are:

{ }
E ( k ) = P ( k ) − P ( k −1) ] / [ I ( k ) − I ( k −1)  =
∆P / ∆I (8)
CE ( k ) =E ( k ) − E ( k −1) =
∆E (9)
where; E denotes the slope of P-I characteristics curve and CE presents the
change of this error; k is the sampling instant; E(k) indicates whether the
operating point at any instant k is situated on the left or right side of MPP
on the P-I characteristics; CE(k) gives the direction of displacement of this
point. The control output from the fuzzy MPPT controller is the change in
duty ratio (∆D) of the DC-DC boost converter. Henceforth, the control action
by the controller is materialized by controlling the change in duty ratio in an
efficient, robust, and smooth manner. The membership functions assigned to
the output variable (∆D) are NB, NS, zero (ZE), PS and PB. The variation
ranges of the inputs are taken from –0.07 to +0.07; whereas the variation
range of the output is –100 to +100 (Figure 17.7).

FIGURE 17.7 Membership functions for: (a) input E; (b) input CE; (c) output ∆D.

17.5.2 INFERENCE ENGINE

Fuzzy inference engine is the tool used for relating the various fuzzified
inputs and outputs with the help of fuzzy rules. This mapping provides a
set of rules by which decisions are taken about the output values of the
controller based on the controller input values. The procedure of fuzzy
A Classy Fuzzy MPPT Controller for Standalone PV System 247

inference encompasses all the linguistic variables used for the inputs and
output mentioned in the fuzzification section along with logical operations in
between the different inputs and if-then rules. Table 17.3 presents the fuzzy
rules used for this proposed work. There are 25 rules altogether involving E,
CE, and ∆D.

TABLE 17.3 Fuzzy Rule Table

CE →
E↓ NB NS ZO PS PB
NB ZO ZO NB NB NB
NS ZO ZO NS NS NS
ZO NS ZO ZO ZO PS
PS PS PS PS ZO ZO
PB PB PB PB ZO ZO

17.5.3 DE-FUZZIFICATION

The procedure of converting fuzzy values into crisp values is called

de-fuzzification. Defuzzification is comprehended by a decision-making
criterion that chooses the best crisp value based on a fuzzy set described
in terms of membership in fuzzy sets. There are plentiful defuzzification
methods including mean of maximum (MOM), center of gravity (COG) and
center average methods. The MOM method can be viewed as the method
which returns the point of balance on a curve, whereas the COG method
yields the value of the center of area under the curve. Amongst those several
methods, centroid method is the most predominant one. In the proposed
work, the COG method is used for defuzzification. Centroid defuzzification
returns the COG of the fuzzy set for the output variable along the x-axis. The
centroid is computed using Eqn. (9), where μ(∆Di) is the membership value
for point ∆Di in the universe of discourse. After the defuzzification, the ∆D
is scaled by S∆D and finally converted into actual duty ratio by the formula
mentioned in Eqn. (10).
n
∑ µ ( ∆D ) .∆D
∆D = i=1 i i
(10)
n
∑ µ ( ∆D ) i=1 i

D ( k=
) D ( k − 1) + S ∆D .∆D ( k ) (11)
248 Advances in Data Science and Computing Technology

17.6 SIMULATION RESULTS AND DISCUSSIONS

The simulation results are tabulated in table IV for different irradiance values
keeping temperature fixed at 25°C. Table 17.4 presents the voltage, current,
and maximum power values from the Fuzzy MPPT controller integrated
system for different temperatures, keeping the irradiance fixed at 1,000 W/
m2 (Figures 17.8 and 17.9).

FIGURE 17.8 Simulink model of (a) the entire system; (b) the DC-DC boost converter.

FIGURE 17.9 (a) fuzzy rule viewer; (b) fuzzy surface.

A Classy Fuzzy MPPT Controller for Standalone PV System 249

TABLE 17.4 The Voltage, Current, and Power Outputs from the Boost Converter for
Different Irradiances Under Standard Temperature 25°C
Temperature Irradiance Boost Converter Boost Converter Power Delivered by
(°C) (W/m2) Output Voltage Output Current the Boost Converter to
(V) (A) the ESU (kW)
25 1,000 442.9 336.7 149.124
35 1,000 441.6 310 136.896
45 1,000 439.7 267.8 117.751

Comparing the simulation results tabulated in Tables 17.4 and 17.5 with
the maximum power point values Vmp, Imp, and Pmp obtained from the PV
module characteristics for different input parameter conditions (irradiance
and temperature) mentioned in Figures 17.4(a) and 17.4(b), it is observed
that the fuzzy MPPT controller is capable of tracking the maximum power
point very closely. The results obtained in this proposed work, is compared
with the results of Hua and Shen (1997) and found to be far better in terms
of transient response of the boost converter’s output voltage and current. The
steady state for voltage is achieved in 0.005 seconds and the currents settle
down in 0.009 seconds. In Oulad-Abbou et al., (2018), P&O method is used.
The voltage settlement is achieved in 0.05 seconds. So, the proposed fuzzy
MPPT controller can track the maximum power point accurately as well as
very rapidly with a sudden change in parameters (Figure 17.10).

TABLE 17.5 The Voltage, Current, and Power Outputs from the Boost Converter for
Different Temperatures Under Standard Irradiance 100 W/m2
Irradiance Temperature Boost Converter Boost Converter Power Delivered by
(W/m2) (°C) Output Voltage Output Current the Boost Converter
(V) (A) to the ESU (kW)
1,000 25 442.9 336.7 149.124
500 25 435.3 174.4 75.916
250 25 431.3 87.11 37.57

17.7 CONCLUSION

The proposed fuzzy based MPPT controller is facilitating precise control

of the duty cycle of the DC-DC boost converter, thus dragging the system
output very close to the maximum power point for the changed values of the
irradiance or temperature.
250 Advances in Data Science and Computing Technology

FIGURE 17.10 Transient response of boost converter output voltage and current.

Thus, it can be concluded that, this controller is an efficient one to put a

fine control in the domain of MPPT. Moreover, due to the non-requirement
of exact system mathematical model to design and implement this controller
along with its robust nature, it can be designed and implemented very fast
and can be used for a very large and complicated system.

17.8 FUTURE SCOPE

The proposed work can be extended by expanding the system. Integration of

EVs charging ports with the existing system for charging the EVs directly
from the output of the DC-DC boost converter can be done. The ESU unit and
the EV charging ports will be connected in parallel. The charging strategy
of EVs along with charging of ESU unit can be designed for proper energy
management of the entire system.

KEYWORDS

• DC-DC boost converter

• duty cycle
• fuzzy logic controller (FLC)
• maximum power point tracking (MPPT)
A Classy Fuzzy MPPT Controller for Standalone PV System 251

REFERENCES

Bendib, B., Krim, F., Belmili, H., Almi, M. F., & Boulouma, S., (2014). Advanced Fuzzy
MPPT Controller for a stand-alone PV system. Energy Procedia, 50(2014), 383–392.
Bhatnagar, P., & Nema, R. K., (2013). Maximum power point tracking control techniques:
State-of-the-art in photovoltaic applications. Renewable and Sustainable Energy Reviews,
23, 224–241.
Coelho, R. F., Concer, F. M., & Martins, D. C., (2010). A simplified analysis of DC-DC converters
applied as maximum power point tracker in photovoltaic systems. In: The 2nd International
Symposium on Power Electronics for Distributed Generation Systems (pp. 29–34). IEEE.
Esram, T., & Chapman, P. L., (2007). Comparison of photovoltaic array maximum power
point tracking techniques. IEEE Transactions on Energy Conversion, 22(2), 439–449.
Hua, C., & Shen, C., (1997). Control of DC/DC converters for solar energy system with
maximum power tracking. In: Proceedings of the IECON’97 23rd International Conference
on Industrial Electronics, Control, and Instrumentation (Cat. No. 97CH36066) (Vol. 2, pp.
827–832). IEEE.
Mirbagheri, S. Z., Mekhilef, S., & Mirhassani, S. M., (2013). MPPT with Inc. Cond method
using conventional interleaved boost converter. Energy Procedia, 42, 24–32.
Nafeh, A. E. S. A., (2011). An effective and safe charging algorithm for lead-acid batteries in
PV systems. International Journal of Energy Research, 35(8), 733–740.
Oulad-Abbou, D., Doubabi, S., Rachid, A., García-Triviño, P., Fernández-Ramírez, L. M.,
Fernández-Ramírez, C. A., & Sarrias-Mena, R., (2018). Combined control of MPPT, output
voltage regulation and capacitors voltage balance for three-level DC/DC boost converter
in PV-EV charging stations. In: 2018 International Symposium on Power Electronics,
Electrical Drives, Automation and Motion (SPEEDAM) (pp. 372–376). IEEE.
Rekioua, D., Achour, A. Y., & Rekioua, T., (2013). Tracking power photovoltaic system with
sliding mode control strategy. Energy Procedia, 36, 219–230.
Salameh, Z. M., Casacca, M. A., & Lynch, W. A., (1992). A mathematical model for lead-acid
batteries. IEEE Transactions on Energy Conversion, 7(1), 93–98.
Salas, V., Olias, E., Barrado, A., & Lazaro, A., (2006). Review of the maximum power point
tracking algorithms for stand-alone photovoltaic systems. Solar Energy Materials and Solar
Cells, 90(11), 1555–1578.
Seddik, M., Zouggar, S., Ouchbel, T., Oukili, M., Rabhi, A., Aziz, A., & Elhafyani, M. L.,
(2010). A stand-alone system energy hybrid combining wind and photovoltaic with voltage
control (feedback loop voltage). International Renewable Energy Congress (pp. 227–232).
Tunisia.
Villalva, M. G., Gazoli, J. R., & Ruppert, F. E., (2009). Comprehensive approach to modeling
and simulation of photovoltaic arrays. IEEE Transactions on Power Electronics, 24(5),
1198–1208.
Zegaoui, A., Aillerie, M., Petit, P., Sawicki, J. P., Jaafar, A., Salame, C., & Charles, J. P.,
(2011). Comparison of two common maximum power point trackers by simulating of PV
generators. Energy Procedia, 6, 678–687.
CHAPTER 18

ACTIVE AND REACTIVE POWER

CONTROL IN VARIABLE WIND
TURBINE USING SIMULTANEOUS
WORKING OF STATCOM AND PITCH
CONTROL
ARRIK KHANNA,1 PUSHPANJALI SINGH BISHT,2 and
VIKRAM KUMAR SAXENA3
1
Department of Electrical Engineering, Chitkara University,
Punjab–140401, India, E-mail: [email protected]
Department of Electrical Engineering, DPGITM Gurugram,
2

Haryana–122001, India
3
Department of Electrical Electronics and Engineering, Pailan College
of Management and Technology, West Bengal–700102, India

ABSTRACT

The energy need is ever attention requiring issue for the under developed and
developing countries. These countries are in the figurative of the shortfall
of hydrocarbons or fossils energy. Wind farms connected to existing power
system poses credible challenges to economic operation of power system.
Wind power is undergoing through a very expeditious development. The
assimilation of wind energy in the existing power system is a concern to
revamp the vitalization of the ability and to maintain the increasing rate of
wind generation installation capacity, this is utmost important in achieving
254 Advances in Data Science and Computing Technology

the set parameters of guarantee and tenability of uninterrupted supply.

Recently, variable-speed wind generators added in wind farm has expanded
and more manufacturers of wind generators are assembling “variable-speed”
wind generators. Adoption of micro grids has added advantage of including
wind generators working with low voltage ride through (LVRT) to stay inter
connected when the voltage and frequency drops. Tripping of wind turbines
generators in fault conditions has a drastic effect on the stability of power
system. This chapter lays light on working of variable-speed wind generators
with STATCOM and pitch control. Analysis obtained from simulation
shows that wind generators can be used at its optimal energy while using
simultaneous Pitch Control and STATCOM for a wide range of wind speeds.

18.1 INTRODUCTION

The evolution in wind energy technology has been exceptional. The cost of
electricity generation from wind energy has decayed to about one seventh
from the early 1980s. In United States, a substantial growth has been seen
in wind power generation capacity, with growth averaging nearly 24% every
year during the past 5 years. In India, as on 31st December 2018, the total
installed wind power capacity was 34,046 MW with generation of 52,666
GWh. But as on 31st December 2019 in India the installed capacity of wind
power has raised to 36,625 MW with generation reaching 62,036 GWh, a
subsequent rise of 7.57% in 12 months. In order to expedite the application
of new wind power projects, numerous computer algorithms using MATLAB
and Homer Pro have been developed for designing and optimization of solar,
wind power, microgrids, and other renewable energy sources. But the major
three challenges to adoption to wind power’s future remains’ the problems
of grid reliability, voltage stability and wind intermittency (Eisa, 2019).
Electrical power generated from wind power has increased exponentially
in global market wind turbines are being equipped with a variable pitch
technology from speed control and static compensator (STATCOM). This
technology is adopted to extract the maximal energy potential and STATCOM
provides stable generation at the grid frequency when generators in wind
turbines operate on the basis of variable rotation speed (Awad et al., 2014;
Fu et al., 2017). As world electricity consumption is on rise steadily, there
is an everlasting onus on society to flourish and adopt the renewable source
of energy for decreasing world carbon footprint and have positive impact on
global environment. In direction of expanding energy intake, wind power
along with solar/ PV generation is the one the swiftly growing sector. As
Active and Reactive Power Control in Variable Wind 255

per data released by Global Wind Energy Council (GWEC) for year 2019
approximately 651 GW of total wind energy has been installed worldwide
with 10% increase from year 2018. China and United States continues to
be largest on shore wind energy market and accounting almost 60% of new
capacity added in year 2019. The potential impact of COVID-19 globally has
nosedived economy and demand for energy for year 2020 was expected to
be a record year for wind generation and as per GWEC with a forecast of 76
GW to be added as new capacity.

18.2 WIND POWER IN INDIA

The installed capacity worldwide for wind power has reached 651 GW by
the end of 2019. World’s top five markets for new wind installations were
China, United States, United Kingdom, India, and Spain (Awad et al., 2014).
The expansion of wind power in India started in last decade of 1990s and has
unquestionably expanded in the last half a decade. In spite of being a relative
novice to the wind business correlated with United Kingdom or Spain, India
lies fourth biggest equipped country with wind power size in the world. As
on 31st December 2019 wind power stands at 3,665 MW in India, mainly
spanning crosswise Tamil Nadu (9231.77 MW), Gujarat (7203.77 MW),
Maharashtra (4794.13 MW), Karnataka (4753.40 MW), Rajasthan (4299.73
MW), Madhya Pradesh (2519.89 MW), Andhra Pradesh (4077.37 MW),
Kerala (62.50 MW), Telangana (128 MW) and other states (30 MW). It is
predicted that 60 GW of further wind power capacity will be added in India
by 2022. Figure 18.1 show wind map of India for the year 2019.
In order to obtain better flexibility and generate more energy to be injected
in system, wind generators of order of MW range are normally interconnected
simultaneously to form a wind farm. Since past 2 decades the size of wind
turbines has increased now, we have turbines ranging from 8–10 MW with
blade span of 160 m (Mahvash, Taher, and Rahimi, 2017). However, integra­
tion of such large variable speed wind generators with existing power grid,
will increase the energy seepage to an equilibrium above which a compelling
brunt effect will be seen on grids operation (Rashad, Kamel, and Jurado,
2018). Previously many tasks are being laid down by regulatory task force
for system stability (Rashad, Kamel, and Jurado, 2018) and consistently
expanding of this renewable power, transmission system handling operators
(TSO) are compelled to lay down new stipulations on the wind farms for the
well-operation of the power grid network.
256 Advances in Data Science and Computing Technology

FIGURE 18.1 Energy MAPS of India: Installed generation capacity (Central Electricity
Authority, 2019).

18.2.1 INTEGRATION OF WIND POWER WITH EXISTING GRID

NETWORK IN INDIA

Integration of wind turbine with existing electric grid network has an effect
on the power quality of power being delivered to the grid or end users. The
Active and Reactive Power Control in Variable Wind 257

power generated from a wind turbine is full of problems such as active

power, reactive power, variation of voltage, flicker, harmonics, voltage swell,
voltage sag, etc. (Al-Jowder, 2007). In variable speed and fixed-speed wind
generators operation, variation in wind acceleration is imparted as change
in the automated torque (Habibi, Rahimi, Howard, 2017). Throughout the
traditional working, wind generator outturns an endless variable output
power. This power change is mainly result of turbulence, wind shear. Thus,
the network connecting wind turbine with existing grid needs to manage for
such fluctuations. In order to eliminate some effects of fluctuations gener­
ated in wind turbines variable speed turbines are being used these days in
conduction with an induction generator (Eisa, 2019). Some of the problems
encountered by using variable speed on grid network are (Habibi, Rahimi,
and Howard, 2017):
• Change in voltage degree;
• Transient voltages and currents;
• Harmonic changes in the waveforms for AC power.
A STATCOM – Pitch control added supervision technology is recom­
mended in this chapter for developing quality power being managed techni­
cally by both effect of variable frequency, voltage, and to reduce mechanical
stresses developed by variable wind on wind turbine blades (Lin et al., 2019),
so as to increase power quality and reduce stresses on power system, which
will help in integrating wind farms in existing power systems.

18.2.2 PROPORTIONAL CONTROL BASED PITCH ACTUATOR

SYSTEM (CSA)

A MATLAB based Simulink model (Oh et al., 2014) is proposed for the CSA
system is shown in Figure 18.2. Wr1 is taken as input which is run through a
congestion filter. The purpose of the congestion filter is to cap the pitch input
in the range of 0 to 90° the normal range of valid pitch angle. Second model
shown below is the gain block which is used to define the proportional gain
of the system, denoted by K-Block 0-pitch_max is also a congestion filter
which is used to define the values of pitching speed. The range of maximum
and minimum permissible values of pitching speed in this system are set at
3.5° per unit time and –3.5° per unit time.
The feedback of this model was realized and noted at three changes
values variable speed matrix. In entire cases pitch angle remained constant
at 75° irrespective of speed and the same is shown in Figure 18.3.
258 Advances in Data Science and Computing Technology

FIGURE 18.2 Block diagram of proportional control based pitch actuator system.

FIGURE 18.3 Pitch angle response of proportional control-based pitch actuator system.

The effect of Pitch angle control on rotor speed, Electrical torque Te and
Mechanical Torque Tm as shown in Figures 18.4 and 18.5.

18.3 DRIVE TRAIN MODEL

The dynamic equations of the two-mass drive-train written on the generator

side are (Habibi, Rahimi, and Howard, 2017):
′
d ωwt
′ J wt
T=
wt ′ ′ − ω gen ) + K (θ wt
+ D(ωwt ′ − θ gen ) (1)
dt
d ω gen
−Tgen J gen
= ′ ) + K (θ gen − θ wt
+ D(ω gen − ωwt ′ ) (2)
dt
Active and Reactive Power Control in Variable Wind 259

FIGURE 18.4 Depicting effect of using constant pitch control on Te and Tm.

FIGURE 18.5 Pitch angle response of pitch control on Vab, Iline, Vab_rms, Iline_rms, Pac, and rotor
speed Wm w.r.t time on x axis.
260 Advances in Data Science and Computing Technology

Here Twt is wind turbine torque, Jwtr is wind turbine moment of inertia,
ῳwtr is spring constant Tgen is generator torque, Jgen is generator moment of
inertia, and ῳgen is spring constant indicating the stiffness of the shaft on
generator part. The torque and is transmitted via the gearbox with a gear ratio
of Ng and angular speed is given as:
′
dθ wt
′ =
Rotor speed ,ωwt (3)
dt
dθ gen
Generator speed ,ω gen = (4)
dt
Figure 18.6 depicts the block diagram of turbine and shaft 2 mass model,
in which the mechanical stress on turbine blades are reduced (Li et al., 2018;
Vidal et al., 2012) with change in speed and the result for which is shown in
Figure 18.7. But with reduction in stress on blades the output shaft torque in
pu is kept almost constant as shown in Figure 18.8.

FIGURE 18.6 Turbine and shaft 2 mass model.

The effect of Electrical Torque Te, Mechanical Torque Tm and rotor speed
Wm using constant pitch control actuator is shown in Figure 18.8. The results
show that after using pitch control technique Te, Tm, Wm remains constant
irrespective of the variable speed and hence providing a Smooth power
generation at variable speed.

18.4 STATCOM (STATIC COMPENSATOR)

Figure 18.9 represents block diagram for AC-DC-AC conversion being used
in wind power generation.
Active and Reactive Power Control in Variable Wind 261

FIGURE 18.7 (a) Reduction of mechanical stresses on turbine blades w.r.t time in sec on x

axis; and (b) constant output shaft torque w.r.t time in sec on x axis.

FIGURE 18.8 Effect of pitch control (x axis) on Te, Tm, Wm. on Y axis.
262 Advances in Data Science and Computing Technology

FIGURE 18.9 Simulink block diagram of AC-DC-AC conversion.

In this block diagram a three phase (two winding) transformer is

followed by a rectifier unit. The basic principle of working is to convert
generated AC-DC-AC with the output AC having a constant frequency
and active reactive power control (Liu, Zhao, and Li, 2014). The rectifier
unit is followed by a LC filter used to eliminate AC waveform present
after rectification; the LC filter network gives its input to PWM IGBT
inverter, which is used to convert DC-AC having a snubber resistance
of 5,000 ohms. The firing angle control of IGBT is provided by discrete
PWM generator.
The basic block diagram of STATCOM is represented in Figure 18.10.
This shows the actual working of STATCOM circuit. A three-phase parallel
RLC link of 50 KW is provided which acts as load for inverted voltage after
AC-DC-AC conversion (Figures 18.11 and 18.12).
Figure 18.13 shows the behavior converted DC voltage, firing angle
control, inverted AC voltage and inverted AC load voltage after using
STATCOM. The figure clearly depicts that one gets near sinusoidal wave­
form of load using STATCOM.
Figure 18.13 shows the three phase power being fed to grid after using
STATCOM. Results obtained through simulation have shown that STATCOM
can drastically raise the restoration of the wind generators because of its
brisk recovery of voltage and hence increased stability when compared with
traditional voltage control techniques. Therefore, it is an appropriate opera­
tion for wind farm business houses to adopt in wind farms for the objective
of engaging the LVRT obligation (Figure 18.14).
Active and Reactive Power Control in Variable Wind 263

Basic Simulink block diagram of STATCOM.

FIGURE 18.10
264 Advances in Data Science and Computing Technology

FIGURE 18.11 Active and reactive power control without using STATCOM.

FIGURE 18.12 Active and reactive power control using STATCOM.

18.4.1 SIMULTANEOUS WORKING OF STATCOM AND PITCH

CONTROL

For efficient working of a wind turbine simultaneous working a PITCH

control and STATCOM has been used in variable speed wind turbine. Here in
Figure 18.15 shows the basic block diagram in which simultaneous working
of PITCH control and STATCOM being implemented.
Active and Reactive Power Control in Variable Wind 265

FIGURE 18.13 Vdc, firing angle, Vab_inv, and Vab_load waveforms.

FIGURE 18.14 Three phase power being fed to grid after using STATCOM.

The problem with using only Pitch control is that as the stress on turbine
blades is reduced considerably along with smooth operation of Turbine
Generator set. But Pitch control has no effect on power quality issues.
Similarly, by using only STATCOM we do obtain a near sinusoidal wave­
form which can be fed to grid at varying speed having constant frequency
 266

FIGURE 18.15 Block diagram of pitch control and STATCOM.

Advances in Data Science and Computing Technology
Active and Reactive Power Control in Variable Wind 267

of operation and providing a quality power (Al-Jowder, 2007) with fewer

burdens on transmission lines and reduced losses (Vidal et al., 2012). As
Figure 18.16 shows the simultaneous working of both Pitch Control and
STATCOM we get a turbine generator set which works on power quality and
less stresses on turbine blades at variable wind speed.

FIGURE 18.16 Effect of active reactive power of turbine using pitch control and STATCOM.

In Figure 18.17 we get the result which depicts the three phase power before
being fed with Pitch Control without load and similarly Figure 18.21 shows
the power being fed to grid after using both Pitch Control and STATCOM
under 50 KW parallel RLC mask link load (Figures 18.17–18.22).
The power generated without using STATCOM and PITCH CONTROL
is 5,816 KW and the power generated using PITCH CONTROL and
STATCOM is 5340.41 KW.
Now ἠ (efficiency) = Output/Input
Hence ἠ in comparison with STATCOM and Pitch control are not being
used
= 5340.41 / 5816 = 91.81%
Which clearly shows that by simultaneous usage of STATCOM and
PITCH CONTROL has very less impact over power loss and has a high
efficiency of 91.81%.
268 Advances in Data Science and Computing Technology

FIGURE 18.17 Three phase power with pitch control without load.

FIGURE 18.18 Three phase power after using STATCOM and pitch control on load.

18.5 CONCLUSION

As the issues that arise due to voltage disturbances, THD, and transients
does not allow power generated by variable speed wind turbine to be directly
Active and Reactive Power Control in Variable Wind 269

fed to grid, hence an intermediate has to be applied so as to eliminate the

effect of THD, transients, voltage swell. This is attained by using PITCH
CONTROL and STATCOM.

FIGURE 18.19 Vdc, firing angle, Vab_inv, and Vab_load waveforms after using both
STATCOM and pitch control technique.

FIGURE 18.20 Active and reactive power being fed to STATCOM using PITCH control.
270 Advances in Data Science and Computing Technology

FIGURE 18.21 Compares the waveform of three phase active and reactive power using
both pitch control and STATCOM.

FIGURE 18.22 (a) Power generated without STATCOM; (b) power generated using
STATCOM and PITCH control.

On comparing Figure 18.15 and 22.22, we see that on simultaneous

working of STATCOM and Pitch Control we get a better controlled and
smooth waveform. Similarly, on comparing phase voltage Vab, Line current
Il, RMS voltage Vab_rms and Pac from Figures 18.15 and 18.22 we get a wave-
form with better frequency control and free from THD and transients.
On studying Figure 18.20 we see that we get a near pure sinusoidal
waveform on no load on using PITCH control, but our main goal was to
use both PITCH Control and STATCOM and then obtain a waveform on
load having reactive power control, which being free from THD, transients,
and high on power quality. According to Figure 18.21 when STATCOM is
introduced to PITCH CONTROL technique and 50 KW RLC load is applied,
Active and Reactive Power Control in Variable Wind 271

we obtain a waveform which is free from ripple/sag/voltage swell effect.

Free from transients and THD. When the voltage or power as shown in
Figure 18.21 is fed to grid then the problem of power quality is minimized.
Similarly, electric devices will work properly without the problem of abrupt/
abnormal overheating, insulation breakdown. This will result in increased
life of electric devices, transmission lines and stress on existing Grid network
will be reduced.
Hence by using STATCOM and PITCH CONTROL simultaneously we
get a better-quality power which can be fed directly to grid. Therefore, instead
of feeding 5,816 KW power, we will feed 5340.41 KW power rich in quality
power. Moreover, the efficiency of simultaneous working of STATCOM and
PITCH CONTROL is very high in order of 91.81% when compared with
model without using STATCOM and pitch control simultaneously. Also,
reduction on power is easily compensated with increase in quality of power
and reduction in THD and transients.

KEYWORDS

• pitch control
• STATCOM
• variable speed

REFERENCES

Al-Jowder, F., (2007). “Improvement of synchronizing power and damping power by means
of SSSC and STATCOM.” Electric Power Systems Research, 77, 1112–1117.
Awad, A. S. A., Shatshat, R. E., Salama, M. M. A., & El-Fouly, T. H. M., (2014). “Low voltage
ride through capability enhancement of wind farms’ generators: DVR versus STATCOM.”
Presented at the 2014 IEEE PES General Meeting | Conference & Exposition.
Central Electricity Authority, (2019). Available: http://cea.nic.in/ (accessed on 08 December
2021).
Eisa, S. A., (2019). “Modeling dynamics and control of type-3 DFIG wind turbines: Stability,
Q Droop function, control limits and extreme scenarios simulation.” Electric Power Systems
Research, 166, 29–42.
Fu, Y., Zhang, X., Hei, Y., & Wang, H., (2017). “Active participation of variable speed wind
turbine in inertial and primary frequency regulations.” Electric Power Systems Research,
147, 174–184.
272 Advances in Data Science and Computing Technology

Habibi, H., Rahimi, N. H., & Howard, I., (2017). “Power maximization of variable-speed
variable-pitch wind turbines using passive adaptive neural fault tolerant control.” Frontiers
of Mechanical Engineering, 12, 377–388.
Li, C., Xiao, Y., Xu, Y. L., Peng, Y. X., Hu, G., & Zhu, S., (2018). “Optimization of blade pitch
in H-rotor vertical axis wind turbines through computational fluid dynamics simulations.”
Applied Energy, 212, 1107–1125.
Lin, Z., Chen, Z., Liu, J., & Wu, Q., (2019). “Coordinated mechanical loads and power
optimization of wind energy conversion systems with variable-weight model predictive
control strategy.” Applied Energy, 236, 307–317.
Liu, Z. J., Zhao, Z. Q., & Li, B., (2014). “Study on application of STATCOM in dynamic
reactive power compensation of wind farm integration.” Applied Mechanics and Materials,
672–674, 255–261.
Mahvash, H., Taher, S. A., & Rahimi, M., (2017). “A new approach for power quality
improvement of DFIG based wind farms connected to weak utility grid.” Ain Shams
Engineering Journal, 8, 415–430.
Oh, K. Y., Lee, J. K., Bang, H. J., Park, J. Y., Lee, J. S., & Epureanu, B. I., (2014). “Development
of a 20-kW wind turbine simulator with similarities to a 3 MW wind turbine.” Renewable
Energy, 62, 379–387.
Rashad, A., Kamel, S., & Jurado, F., (2018). “Stability improvement of power systems
connected with developed wind farms using SSSC controller.” Ain Shams Engineering
Journal, 9, 2767–2779.
Vidal, Y., Acho, L., Luo, N., Zapateiro, M., & Pozo, F., (2012). “Power Control Design for
Variable-Speed Wind Turbines.” Energies, 5, 3033–3050.
 PART VI

Technique for Improving Information and

Network Security

CHAPTER 19

INFORMATION SECURITY USING KEY

MANAGEMENT
SANJIB HALDER,1 BIJOY KUMAR MANDAL,2 and
ARINDAM BISWAS3
1
The Bhawanipur Education Society College, Kolkata, West Bengal,
India
2
NSHM Knowledge Campus, Durgapur, West Bengal, India,
E-mail: [email protected]
3
Kazi Nazrul University, Asansol, West Bengal, India

ABSTRACT

To exchange secret messages between two users, while using the processes
of encryption and decryption on the message with the help of a secret key.
We will see briefly how the key is being generated. Key will be generated
by the process of Diffie-Hellman key exchange algorithm, which does not
follow the rules of sharing secret key instead two parties will develop the
secret key together while exchanging partial information for key generation.
After the generation of secret key, we will use DES (data encryption stan­
dard) algorithm for sending the message, while using the public media, i.e.,
Internet as communication channel and in receiver’s side show that message
is well received without any third party alteration and can be decrypted
perfectly into originally sent plain text.

Advances in Data Science and Computing Technology: Methodology and Applications. Suman Ghosal,
Amitava Choudhury, Vikram Kumar Saxena, Arindam Biswas, & Prasenjit Chatterjee (Eds.)
© 2023 Apple Academic Press, Inc. Co-published with CRC Press (Taylor & Francis)
276 Advances in Data Science and Computing Technology

19.1 INTRODUCTION

We human beings always tend to keep privacy among us, whether it is about
precious treasure or day to day life chores, or our intensions (Hellman and
Diffie, 1976; Lin and Wang, 2013). To keep it secret we tend to use means
of secrecy, some secret signs that will prevent unwanted public unaware of
what’s going on. Here comes the use of cryptography (Eric and Chachati,
2014). Around 1900 B.C. Egyptians used non-standard hieroglyphs to
inscribe some documents that is the first documented use of cryptography.
Cryptography is nothing but an art of hiding secret in plain-site, so that even
if it’s visible deciphering the meaning behind is impossible (Pervaiz, Cardei,
and Wu, 2010). Scientifically saying it conveys the message from sender to
receiver, while restricting threat data theft or unauthorized access. To keep
the information safe cryptography involves generating codes, it may be in
written form or generated by machine. Data is converted into a format that is
unreadable for unauthorized user, only allowing it to be transmitted without
compromising the data, means restricting decoding it back into a readable
format by unauthorized entities (Ballardin, 2011; Carmen, 2012). Modern
cryptology exists at a junction of subjects that include the use of various
types of rules and discipline like physics, mathematics, communication
science, computer science, electrical science. In addition, its application is
spread across several fields like military communications, business commu­
nications, digital currencies/ transactions, electronic commerce, chip-based
payment cards, etc. Cryptography is used on several levels for Information
security to fulfill distinct tasks (Chen, 2013; He and Lee, 2008).
There are five primary functions of cryptography nowadays:
1. Confidentiality: It ensures that except for intended and authorized
user no other person can access the data or read the message.
2. Key Exchange: The process, which involves the exchange of keys
between sender and receiver.
3. Integrity: It assures the receiver that no kind of alteration or changes
have occurred in the message, and the message has reached in its
original form.
4. Authentication: It is the process of proving one’s identity, i.e., the
message is sent by an intended sender also intended receiver receives
it. No third party has accessed the message.
5. Non-Repudiation: It is a mechanism to prove that the sender’s
involvement. When a message is sent and reaches the receiver,
during cross checking sender may deny the report of having sent the
Information Security Using Key Management 277

message, here non-repudiation comes to work. It is like a signature,

which cannot be denied.
Cryptography consists of few simple steps. We start with the unencrypted
data, also known as the plain text. Plain text is readable format of the message,
which is easily understandable by anyone (Liaw et al., 2005). Now using the
secret key, we change the unencrypted date to encrypted form called cipher
text. The cipher text is sent from sender to receiver through some channels of
communication. Receiver will decrypt the coded message into simple plain
text. Cryptography has always been a great means for intelligence gathering
(Saeed and Shahriar, 2011). Lastly, cryptography is most closely related with
the creation and development of mathematical algorithms, which are used
in the process of encryption and decryption of messages. It is the science
of analyzing and breaking encryption scheme, also use the deciphered
information for greater good of society (Fan, Wen, and Zhang, 2010).
Key Exchange is the process of sharing cryptographic keys between
two parties, sender, and receiver prior to exchange of messages, using
cryptographic algorithm (Xu, Zhu, and Feng, 2011). As when sender and
receiver wish to exchange secret messages, they must be equipped with keys
in order to encrypt the message to be sent and to decrypt the message the
received. In-band and out-of-band are the two ways in which key exchange
can takes place. FTPS, HTTPS, and SFTP, etc., several secure file transfer
protocols have to encrypt the data through symmetric encryption to preserve
the confidentiality during data transmission (Ali, Sivaraman, and Ostry,
2013; He, Niedermeier, and De Meer, 2013). This type of communication
requires both the parties to have the same encryption key. In the world two
communicating parties may be globally situated at end two ends of the earth,
hence physically meet up for every time they need to do a transaction is not
at all convenient. To make life easier we use public communication channels
like Internet to do the talk, but data theft can happen anytime, as it is a
public server. Here comes the need of secret means of communication (Xu,
Zhu, and Feng, 2011). In addition, eventually the need of key exchange that
will ensure later send messages do not fall in unauthorized hands. FTPS
and HTTPS file transfer protocols which are SSL/TLS protected uses the
key exchange occurs by the process of handshake. It works in request
basis, where usually a web browser or a client requests a connection to the
intended server by sending message, known as Client Hello. Client Hello
consists of some random data like session id, compression algorithms and
mainly a cipher suite. When the server allows the permission, a connection is
established. Each cipher suite will have separate algorithm for key exchange,
278 Advances in Data Science and Computing Technology

encryption, and message authentication. As the server receives, the Client

Hello it searches for supported cipher suite in its own set of lists, after
comparison choose the best cipher suite, and eventually choose the desired
key exchange algorithm. In addition, immediately set to exchange key using
the chosen key exchange algorithm, and proceed to share messages.
The two most popular key exchange algorithms are RSA and DIFFIE­
HELLMAN. It probably would not be too much of a stretch to say that the
advancement of these two key exchange protocols accelerated the growth of
the Internet. That is because these two protocols allowed clients and server,
as well as server and server, to exchange cryptographic keys over insecure
medium (Internet) and in turn enable them to transact electronically in a
secure manner.

19.2 PREVIOUS WORK

The first invention got the root of key exchange was in the hand of Ralph
Merkle who developed the ‘public key agreement’ technique in 1974. Later,
it became popular as ‘Merkle’s Puzzles’ and got published in 1978. Before
then, in 1977, Ron Rwest, AdiShamia, and Leonardo Adleman from MIT,
discovered a Cock’s Scheme autonomously. The well-known protocol was
proposed by Bailey Whitfield Diffie, an American cryptographer, with
collaboration of Martin Hellman and Ralph Merkle, who became a spearhead
in development of public-key cryptography. The concept of Diffie-Hellman
was published in the journal ‘New Directions in Cryptography’ in 1976
which had set a new field in distribution of cryptographic keys over an in
secured communication channel. The problem for this key distribution was
solved using this public key cryptography algorithm and communication
became more secured then before. This also initiated the root of another
advanced encryption algorithm, which is asymmetric key algorithm.
Since, Merkle the key exchanging phenomenon, Hellman suggested
that the algorithm should be called Diffie-Hellman-Merkle’s algorithm in
2002 set the first footmark. The AKE protocol is also called STS protocol
developed by Diffie in 1992. This is an advancement of raw Diffie-
Hellman Key exchange protocol because it authenticates the two parties in
communication. The mechanism follows as, the two parties can calculate a
secret key using Diffie-Hellman algorithm and then they get authenticated
by the exchange of digital signatures between them. The STS protocols
prevents the man-in-middle attack due to authentication. It provides the
forward secrecy with the help of Diffie-Hellman algorithm to compute the
Information Security Using Key Management 279

session key that is short termed. It also provides the facility of unaffected
session of previous communications using the long term keys. It shortens the
process of communication, as the central authority need not be interfered due
to authentication and prevents redundant exchange of elements.

19.3 PROPOSED ALGORITHM

The Diffie-Hellman algorithm is being used to establish a shared secret that

can be used for secret communications while exchanging data over a public
network. We need to remember the point that we are not sharing secret,
instead creating path that will be used later as communicating channel.
Few points to remember before writing the algorithm are as follows:
 For the simplification and practical and easy implementation of
the algorithm, we will use only 4 variables one prime number P, M
(a primitive root of P) generator and two private values of two parties’
c and d.
 M and P are publicly available numbers. Users (say Alice and Bob)
pick private values c and d and they generate a key and exchange it
publicly. The other party receives the key and from that generates a
secret key after which both parties will possess same secret key for
encryption and decryption.
• Both Alice and Bob set an agreement about the prime number and the
generator P and M where (0<M<P).

e.g., =23 and M=9

• Alice selects a random secret number that is her private key as c and
Bob selects the same as d.

e.g., c=4 and d=3

• With the help of the private key and the other two parameters Alice
and Bob compute their public values
A. AliceX = (M^ c)mod P

X = (9^4) mod 23

B. BobY = (M^ d)mod P

Y = (9^3) mod 23

• After that, Alice, and Bob exchange their public numbers between
each other, i.e., X is sent to Bob and Y is sent to Alice.
280 Advances in Data Science and Computing Technology

• Alice receives public key of Bob as Y, and Bob receives public key
of Alice as X
• Alice and Bob compute symmetric keys to trick the third party.
o Aliceka = (Y^ c) mod Pka=(((9^3)mod23)^4) mod 23
o Bobkb = (X^ d)mod P kb=(((9^4)mod 23)^3) mod 23
• If K = ka = kb, then K is the shared key generated. K=9
• The shared secret key K is transformed to 64 bits and with the plain
text (64 bits) it is given as input to the DES algorithm.
Initial Permutation: This step produces two equal halves of permuted
block, i.e., LPT, and RPT with each having 32 bits and 16 rounds of encryp­
tion process is being carried out on each LPT and RPT.
1. Transformation of Key
i. For the removal of 1 bit from each byte of block for parity
checking, 56 bits are actually available which is the actual key
length to be used for encryption. Here we get this key from
previously calculated Diffie-Hellman key exchange algorithm.
ii. For each round of permutation, a different set of key of 48 bits is
derived from the above length.
iii. 56-bit key is divided into two equal halves, which has 28 bits of
length in each.
iv. Depending on the rounds, the bits are circularly shifted to the
left by one or two positions.
v. After appropriate shift, 48 bits from 56 bits are selected. This is
called Compression Permutation as it compresses the number of
bits.
2. Expansion Permutation
i. The right hand side block RPT is expanded from 32 bits to 48
bits and also, they are permutated.
ii. RPT having 32 bits is divided into 8 blocks of bits each.
iii. The above 4 bits of each block is expanded to a corresponding
6-bit block, i.e., 2 more bits are added which is the repetition of
1st and 4th bit (48 bit, 8 blocks each of 6 bits),
iv. The compressed key of 48 bit from key transformation part is
XORed with 4 RPT of 48 bits from the previous step and the
output is fed to the S-Box subset.
3. S-Box Substitution
i. Substitution that is performed is in 8 substitution boxes.
ii. Each 8 S-boxes have input of 6 bits and output of 4 bits.
Information Security Using Key Management 281

iii. The 48 bit input block is divided into 8 sub blocks and each
block is given an S-box.
iv. The output produced from all the S-Boxes are combined to form
32-bit block which is given to the P-Box permutation as input.
4. P-Box Permutation
i. Simple permutation takes place. Each of the bits are swapped
with each other.
5. XOR and Swap
i. LPT is XORed with output of P-Box.
ii. Result produced in the above step is the new right half.
iii. The old right half now forms the new left half.
a. Final LPT and RPT are recombined to form a single block
and final permutation is performed on the rejoined block
where there is a simple transposition of each bit. This is
performed at the end of all the 16 rounds.
b. Four results that are produced is a 64 bits encrypted cipher
text.
c. To reverse or decrypt from the cipher text to the plain text
or the original text, the reverse of all the above steps are
performed along with the key that is generated in Diffie-
Hellmen key exchanging algorithm.

19.4 WORKFLOW

The above workflow gives a clear view of the combination of the Diffie-
Hellmen key Exchanging Algorithm with DES algorithm. Alice and Bob
agree to a prime number and a generator that is much smaller than the
prime number and has to be co-prime with it. They both has a private
key with which they can calculate their public keys and exchange them
with each other. After the exchange, they calculate their shared key from
the parameters present with them. If both the key matches then only they
can have a communication with each other, otherwise, not. If the keys are
matched, then it is given as an input along with the plaintext to the DES
encryption algorithm where it generates the cipher text after going through
several phases of transformations. The cipher is decrypted using the same
key by DES decryption algorithm (Figure 19.1).
A proper analysis of the above proposed design is being done. The time
taken by both combination algorithm is much larger and hence it prevents a
third party from interrupting.
282 Advances in Data Science and Computing Technology

FIGURE 19.1 Workflow.

19.5 ANALYSIS AND DISCUSSION

A key is nothing but a stream of bits generated using some cryptographic

algorithms. It is the ‘heart’ of cryptosystems. A cryptographic key can be
compared with a real life lock and key. Suppose, we have locked a baggage
using a key and parceled it to another person. The person can only be able
to open the lock if he/she has the same key or a key that is able to open
the particular lock. The cryptographic keys work in the same manner. Two
person communicating with each other has a key with them, which enables
them to encrypt or decrypt a message. For that, either the key is same for
the sender and receiver or there is a paired key, which is different for both
encryption and decryption.
The usage of keys are more than encryption and decryption. They are
also used for digital signatures because each signature has a separate identity
Information Security Using Key Management 283

that is identified by a key. Message Authentication is another utilization of

keys (Figures 19.2 and 19.3).

FIGURE 19.2(a) Key generation.

FIGURE 19.2(b) Key generation of with clients.

FIGURE 19.3(a) Plain text.

FIGURE 19.3(b) Generation of cipher text.

FIGURE 19.3(c) Decrypted message.

284 Advances in Data Science and Computing Technology

The Diffie-Hellman key exchanging algorithm is a key generation and

exchanging algorithm where two parties communicating with each other
share a common secret key between them before the actual communication
process starts. As soon as the communication link is set up, a key is gener­
ated and exchanged by any one of them. The Diffie-Hellman design solves
the purpose. Before the communication the keys are not known to them.
After the interchange of the key, the same is used to encrypt the message.
The cipher text generated is then received by the receiver and he/she uses the
same key to decrypt the message. Hence, the receiver receives the original
message.
The Diffie-Hellmen Key exchanging algorithm follows the mechanism
in the given manner. There are two users namely Alice and Bob. They
would communicate with each other but before that, they need to exchange
a key. Actually, the exchange of key is an illusion. Keys are derived in joint
collaboration, which gives an illusion of exchange. They set an agreement
with two prime numbers one is the P and other is the generator G. The P has
to be quite large in in comparison to the G. The limit should be 0<G<P. Alice
has a private key x and Bob has a private key y. They keep this with them.
Alice calculates (G^x) mod P and Bob calculates (G^y) mod P. Now these
results are exchanged between them, which called public keys. Now Alice
computes (G^y)^x mod P and Bob computes (G^x)^y mod P. For both cases,
the result becomes (G^xy) mod P. This is the shared key. Now if am intruder
knows the values P, G,(G^x) mod P,(G^y) mod P he can figure out x and y
which are there private keys. It can only be possible if P value is small. This
is the reason the value of P is takes a large prime number which leads to a
discrete logarithmic problem. To figure out the private numbers it would be
time taking and by the whole process would be completed.
Now both the parties using DES algorithm use for encryption as well as
for decryption this shared key. The DES has a drawback of having a very
small key length that is 56 bits. But when the key from Diffie-Hellmen is
taken it takes quite a large amount of time and a new key is generated for
every other session. Hence, the key size does not affect the process much and
the communication is done in more secured process.
The same key that is generated is farther given as an input to the DES
algorithm. Since, we are giving an external input to the keys; the algorithm
need not compute the keys using the round function. This reduces the task
of DES algorithm and hence enhance its efficiency. We provide a plaintext
and key of 64 bits where only 56 bits are used in actual and the rest are used
for checking the parity. Therefore, the actual length of the key becomes 56
Information Security Using Key Management 285

bits. Each byte of the key is sent with an odd parity. In each byte.one of
the bit is used for checking the error while generating and distributing the
key. Generally, the 8th, 16th, …, 64th are used for checking the parity. Several
permutations are performed in the plain text with the key to produce the
cipher text. To convert the cipher text to the plain text again the reverse order
is performed.
Hence, this makes the algorithm efficient in comparison to the individual
Diffie-Hellmen and DES algorithms.

19.6 FUTURE SCOPE

Diffie-Hellman is not a symmetric algorithm; it is an asymmetric algorithm

used to establish a shared secret for a symmetric key algorithm. It works
on secret communication while exchanging data over public network using
Elliptical Curve to generate points and get the secret key using the parameters.
A padlock icon is can be seen next to the URL of your device every time you
go to a web browser; it can be seen because that server has used D-H:
1. TSL/SSL: These two layers use both D-H and symmetric encryption
algorithm. D-H was used initially for the creation of the session key
(shared secret) that is used afterwards in message communication for
that particular session.
2. Perfect Forward Secrecy: The algorithm provides a perfect forward
secrecy. After the shared secret key is known, it can be used to encrypt
a traffic of data. Later, even if the data traffic is known it would be
impossible to know the key. The exchange may also be visible but
the key would be irrecoverable as it was never transmitted and saved.
3. Trapdoor Function: This works as a one-way function. This means
you can calculate in one direction but opposite is not possible and
can only be done by the help of some special information and this
works like a “trap.” The Diffie-Hellman design has this disadvantage
for the interceptors.
4. Authentication: Diffie-Hellman design is well enough to work on
symmetric algorithms. The man-in-the-middle attack problem is
solved to some extent when it is combined with the DES algorithm
as it takes a large amount of time ion exchanging the key, encrypting,
and decrypting the messages. However, there is still a question about
the authentication of the two parties. They can be masked if somehow
or the other the prime number and the generators are very small and
286 Advances in Data Science and Computing Technology

can be cracked by the intruder because of easy discrete logarithmic

calculations. For the same reason, this is used with digital signatures.
5. Time Out: This is another problem faced by the algorithm. The
crackers could not crack the keys until now but the time would finish
before the communication is completed. This problem compels NSA
to move to the ECC for farther security.
Small Key Length: Due to the smaller key length, this would not
be secured in the future. Moreover, it has to be applied with some
other algorithms to enhance the key size to increase the security
properties farther.
6. ECDH: There is a high chance of the Elliptic Curve Discrete
Logarithm Problem to gain popularity in the near future since it
operates on a smaller key size but provides a larger security then
normal Diffie-Hellman problem. None of the parameters are open to
anybody except for the public keys.

19.7 CONCLUSION

When the DES algorithm is used for encryption, the key used in it is
provided by some inbuilt function or algorithm within it. This may raise a
question to the security because it can be attacked by an outsider while the
communication is going on. This is the reason DES algorithm is not used
presently in practical life. But when we have combined the Diffie-Hellman
Key Exchanging algorithm with the DES, the key generated is instant. This
means the key was unknown to both the sender and receiver before the
communication link was established. When it is prior unknown to both of
them, it does not permit the intruders to know the keys and identify any of
the two parties, which would allow them to attack the information. Hence,
using the Diffie-Hellman we can create an instant session key and exchange
the keys between the sender and receiver. Once the keys are exchanged, the
messages are sent by the sender, which is encrypted by the DES algorithm
using the key from Diffie-Hellman key exchanging algorithm, i.e., exchanged
between the two communicators. The receiver receives the message and
decrypts the same with the help of the key that is already present with him/
her. Hence, the communication ends here. It is noted that for every time a
link is created between them again, a new key is interchanged which means
even if the interloper tries to know the key it would be useless since for the
next time, he/she has to compute another key for knowing the information.
The session key exchange also signifies that the time taken would be more
Information Security Using Key Management 287

than the normal encryption algorithms. Hence, it would take quite a long
time to break the key because of larger computation. The time the intruder
would break the key the communication would be completed between the
two people. Hence, the combination of the two algorithms would provide a
better security then the other symmetric key encryption algorithms.

KEYWORDS

• data encryption standard (DES)

• SFTP
• SSL
• TLS

REFERENCES

Ali, S. T., Sivaraman, V., & Ostry, D., (2013). Eliminating reconciliation cost in secret key
generation for body-worn health monitoring devices. IEEE Transactions on Mobile
Computing, 13(12), 2763–2776.
Ballardin, F., (2011). A calculus for the analysis of wireless network security protocols.
Formal Aspects of Security and Trust (pp. 206–222). Springer: Berlin Heidelberg.
Carmen, R., (2012). Wireless Network Security. Ovidius University; Annals Economic Science
Series.
Chen, L., (2013). Applications, technologies, and standards in secure wireless networks and
communications. Wireless Network Security (pp. 1–8). Springer: Berlin Heidelberg.
Eric, G., & Chachati, M., (2014). Analyzing routing protocol performance with NCTUns for
vehicular networks. Indian Journal of Science and Technology, 7(9), 3191–1402.
Fan, Y. J., Wen, Q., & Zhang, H., (2010). Smart card-based authenticated key exchange
protocol with CAPTCHA for wireless mobile network. In: 2nd International Conference on
Future Computer and Communication (ICFCC) (pp. 119–123).
He, X., Niedermeier, M., & De Meer, H., (2013). Dynamic key management in wireless sensor
networks: A survey. Journal of Network and Computer Applications, 36(2), 611–622.
He, Y. H., & Lee, M. C., (2008). Towards a secure mutual authentication and key exchange
protocol for mobile communications; In: 6th International Symposium on Modeling and
Optimization in Mobile, Ad Hoc, and Wireless Networks and Workshops (pp. 225–231). Berlin.
Hellman, M. E., & Diffie, W., (1976). New directions in cryptography. IEEE Transactions on
Information Theory, 22(6), 644–654.
Liaw, S. H., Su, P. C., Chang, H. C., Lu, E. H., & Pon, S. F., (2005). Secured key exchange
protocol in wireless mobile ad hoc networks. In: 39th Annual 2005 International Carnahan
Conference on Security Technology (pp. 171–173).
288 Advances in Data Science and Computing Technology

Lin, Q., & Wang, Y., (2013). Novel three-party password-based authenticated key exchange
protocol for wireless sensor networks. Advances in Wireless Sensor Networks, 334, 263–270.
Pervaiz, M. O., Cardei, M., & Wu, J., (2010). Routing security in ad hoc wireless networks.
Network Security (pp. 117–142). Springer, US.
Saeed, M., & Shahriar, S. H., (2011). security analysis and improvement of smart card-based
authenticated key exchange protocol with CAPTCHAs for wireless mobile network. IEEE
Symposium on Computers and Communications (ISCC) (pp. 652–657). Kerkyra.
Xu, J., Zhu, W. T., & Feng, D. G., (2011). An efficient mutual authentication and key agreement
protocol preserving user anonymity in mobile networks. Computer Communications, 34(3),
319–325.
CHAPTER 20

STRENGTH-BASED NOVEL
TECHNIQUE FOR MALICIOUS NODES
ISOLATION
PRACHI CHAUHAN1 and ALOK NEGI2
G.B. Pant University of Agriculture and Technology, Pantnagar,
1

Uttarakhand, India, E-mail:[email protected]

National Institute of Technology, Uttarakhand, India
2

ABSTRACT

The ad-hoc vehicle network is a distributed and self-configuring network

that does not have a central controller. Due to the high mobility of vehicle
nodes, route establishment through source to destination is really the biggest
challenge on the network. Different techniques have recently been studied
for the identification of malicious nodes in any network. The proposed
work is based on monitoring mode and signal strength. The simulation was
conducted on NS2 and experimental results show that the proposed technique
showed descriptive results in terms of different parameters.

20.1 INTRODUCTION

MANET is indeed a mobile ad-hoc network made up of a versatile

infrastructure the whole network is often a comprehensive wireless network
consisting of specified mobile packet nodes within it. After that, whenever
an arbitrary configuration is developed, the connection of nodes occurs to
each other on the random basis. The nodes mostly on network may also
290 Advances in Data Science and Computing Technology

be known as routers or hosts (Camp et al., 2002). The self-configuration

feature of such an infrastructure performed the communication in an efficient
manner throughout the entire network. VANET is a part of the mobile ad-hoc
networks. For research purposes, the vehicle ad-hoc networks seem to be
the most prominent area of interest. For intent of communication several
vehicles and their components (Chauhan et al., 2017) across the roadside
are always linked to one another, so this infrastructure should also be self-
configured in design. They need no static infrastructure for them (Rajamani
et al., 2000). The routing structure of the network is modified whenever the
system topology is disrupted, however, there are widely changing nodes
or vehicles participating in the system generate very difficult situation
to perform tasks. Again, for aim of forwarding packets between nodes, a
greedy position-based routing technique known as edge node based greedy
routing (EBGR) is often used. There are some nodes accessible on the edge
of the source or forwarding node transmission range. Based on the potential
score of its nearest node, the much more appropriate next hop is appointed
(Mustafa et al., 2010). There is a minimization of the end-to-end latency
of the packet propagation as compared with the existing VANET routing
protocols. Protection of VANETs is violated by numerous forms of attack.
The attacks happen once when one node keeps causation multiple messages
to alternative nodes that are assumed to be from completely Varying
identities. In certain instances, a Sybil attack always be feasible. Attacker
pretends to be having several identities or nodes throughout this attack. A
malicious node may behave as a real node even though it was more than
one number of nodes, whether by impersonating various nodes or simply by
claiming false identities (Rawat et al., 2012). The collision in the network
begins which ends in inflicting Sybil attack within the network. The primary
goal of DoS attack seems to be prevention of a legitimate user from using the
resources as well as the services. Along with the network, the entire channel
can also be jammed in this attack. This result is an inability of not being
able to access the network by the authorized vehicles. The DDOS attack is
more harmful than the DOS attack, because of its distributive nature. For the
purpose of launching the attack, various types of locations are used. Various
time slots can be used for the purpose of sending the messages where the
natures of the message, as well as the time slot, are different for each vehicle.
V2V and V2I can both have DDOS attack them. The unauthorized users use
the reply attack which helps them to be a legitimate user or the road side
unit (RSU) (Malla et al., 2013). The replaying of already generated frames
is done to the new connection by the attacker. In wormhole attack, a packet
is received by the malicious node in the network at a point which tunneled
Strength-Based Novel Technique for Malicious Nodes Isolation 291

back to another point at another location where rests of the packets present
within the network. Inside VANET, there are different routing protocols. In
DGR (directional greedy routing), Via choosing the nodes that heads nearer
target destination, the hop count reduced (Iqbal et al., 2009). A-STAR stands
for Anchor-based-street and traffic aware routing. It is a position-based
routing mechanism which uses city bus routes for the purpose of identifying
an anchor path for packet delivery along with the high connectivity. GyTAR
(Greedy Traffic-Aware Routing) is the proposed technique which used in
city environments where the intersection kept as the base. The geographical
routing protocol is used for this technique. E-GyTAR is the modified version
of GyTAR routing protocol. The junction is selected dynamically because
it based on vehicular traffic density. The AODV is an on-demand routing
protocol, which described as a reactive routing protocol (Ahmad et al., 2010).
Sometimes throughout the network, the source node needs to be routed to a
particular destination then the routing establishment initialized by the route
discovery process. Every neighbor relay the RREQ packet on their own to
the neighbors of the source node.

20.2 LITERATURE REVIEW

Rajamani et al. (2002) have proposed the spacing policies which help in the
highway vehicle automation. Here, a spacing policy (a nonlinear function of
speed) is discussed. The autonomously present information is utilized by the
spacing policy. The guarantee of proving string stability by the spacing policy
seen after the analytical calculations. Also, it ensures the traffic flow stability
by keeping the smaller steady state spacing. Larger traffic flow capacities
which in range of 20–65 m/h are provided through this mechanism. The
adaptive cruise control (ACC) vehicles present on the highways of today can
readily use the spacing policy.
Peel et al. (2001) have developed a design framework of intelligent trans­
port systems which related to the aspects of road sweeping vehicle automa­
tion. The exchange of information from car to car as well as car to road is the
main aim for the road condition information transferring module. The security
issues of VANETs are discussed and their solutions to be derived. This is been
made sure that solutions implemented under certain security patterns.
Camp et al. (2002) have reviewed mobility models for ad-hoc network.
In this chapter, the researchers recommended security infrastructure for
vehicular communication. The main aim of proposed structural design
holds organization of identities. Further there should be the involvement of
292 Advances in Data Science and Computing Technology

cryptography keys which ensure the security of communications and inte­

grate the privacy enhancement technologies. The parameters which could
be enriched such as to provide improved privacy and security protection for
future that aimed to be achieved by the system proposed.
Wu et al. (2011) have explained a new technology which proposed for
the communications being held at vehicles present within shorter range.
Both the V2V and V2I communications are introduced under the highway
scenario. The work discusses the network characteristics of the driving
network environment through this chapter.
Kwag et al. (2006) have discussed the performance evaluation of the
IEEE 802.11 ad-hoc networks in this chapter. This evaluation is performed in
the vehicle to vehicle (V2V) communication for analyzing the IEEE 802.11
ad-hoc networks performance. The V2V communication which occur in
vehicular background focuses on justification which is important for security
correlated services. The flexibility effects are also considered here.
Haas et al. (2007) have devised performance measurements of simulations
of VANETs. The information traces of vehicle movements which created by
the traffic simulators are depend on traffic model theory. The research that
archives movement of vehicles based on broad scale recordings discussed in
this chapter.

20.3 SYBIL ATTACK IN VANETS

The attack occurs when a single node keeps sending multiple messages to
other nodes which are pretended to be from varying identities. In certain
instances, Sybil attack is possible. It can only be exempted from either the
extreme conditions as well as possibilities of resource parity and entity
coordination. A type of confusion occurs in the whole network when a single
node starts sending multiple copies of its selves. There is a chance that all the
fake, illegal ID’s and the authority are claimed (Figure 20.1).
The collision within the network starts which causing Sybil attack in the
network. Both internal and external attacks can be triggered by this type of
attack. That being said, by establishing metrics of authentication, external
attacks can indeed be prevented. For internal attacks, it is not feasible.
There are one to one mapping of the identity and role inside a network.
The Sybil points represented as A, B, C, D nodes in Figure 20.1, generate
a false identity inside the network. Sybil attack has become such a crucial
attack whereby multiple messages are sent from the attacker and then sent to
various vehicles with multiple IDs every time. This leads all other nodes to
Strength-Based Novel Technique for Malicious Nodes Isolation 293

get distracted in such a manner that the nodes think the messages are coming
from certain nodes. Because of this, there is indeed a jam inside the network.
This causes the vehicle to take another route and leave that path to something
like the advantage for attacker. There are two types of Sybil attacks namely,
sensitive, and throughput sensitive attacks.

FIGURE 20.1 Sybil attack.

20.4 PROPOSED TECHNIQUE

In this work, technique has been discussed which detect and isolate malicious
nodes that accountable to trigger Sybil attack from network. The suggested
approach involves node strength methodology in monitor mode. In this
technique, the RSUs flood ICMP messages in network. The vehicle nodes
when obtain the ICMP messages will begin Trying to send towards nearest
roadside units (rsus) its transmission strength rate All details will be obtained
by the roadside infrastructure and the info will be shared with one another.
The node of vehicle with multiple signal strength values will be detect as the
node which Trigger an intrusion into the networks. The RSUs transmit the
controlling information packets in the network to validate which node would
be the fraudulent node and then once the control packets are received, nodes
go out to monitor mode and continue monitoring their neighboring nodes.
The node that’s suspicious be detected and multiple path routing technique
applied which isolate malicious nodes from the network.

20.4.1 WORKING STEPS FOR PROPOSED WORK

• Firstly, road side unit (RSU) save the information regarding deploy
vehicles;
294 Advances in Data Science and Computing Technology

• If any new vehicle joins the network, then all information (network id,
route, etc.), send to RSU;
• Identity changed by any vehicle send important message to all nearest
RSU;
• RSU check the adjacent vehicles information based on signal strength;
• Finally, if the signal strength matched then communication start in the
network. Otherwise, vehicle nodes go into monitor mode and detect
the illegal nodes.

20.5 RESULTS AND DISCUSSION

The simulation of proposed and existing algorithm is done on NS2 by

considering following parameters and values.

20.5.1 SIMULATION DESIGN

For simulation purpose, we construct a network, which comprises wireless

ad-hoc network for the formation of simulation result. As we all know that
VANET is performed in a real environment, hence for the better analysis
of the results of our protocol we design a realistic scenario of road traffic.
The network, which we configure is consists of vehicles, routes, and
RSUs.
To test the performance of all above present methodologies, a program
is written in TCL and tested in real environment. The snapshots of proposed
work execution are shown in Figure 20.2.

FIGURE 20.2 Initialization of network.

Strength-Based Novel Technique for Malicious Nodes Isolation 295

In this work, at first finds the signal strength of each node, i.e., signal
strength of every node and their neighbors. Figure 20.2 demonstrates the
initialization of network.
In Figure 20.3, vehicular ad hoc network is installed with static number
of road side sensor nodes and vehicles which travels freely on roads and V2V
communication and vehicle to road side communication is also accessible.

FIGURE 20.3 Network deployment.

In Figure 20.4, the smart cars are moving on roads and communicating
with one another, i.e., car with identification number 1 communicating with
car that registration number is 4.

FIGURE 20.4 Communication between cars.

296 Advances in Data Science and Computing Technology

In Figure 20.5, new car gets its registration identity and this information
flooded in the network.

FIGURE 20.5 Allocation of new identity.

In Figure 20.6, newly registered nodes change their identity like 2 to

4 but a node with identity 4 already exists in the network. It means Sybil
attack has been triggered in the network. The mischievous nodes have short
distance from legitimate node. So, source node begins sending information
to malicious nodes.
In Figure 20.7, the RSUs’ gathering the information of signal strength of
each car after Sybil attack triggered. As in network two cars of ID number 4
exits and signals are different. It means there are certain malicious cars exist
in network.
In Figure 20.8, to detect particular malicious cars from network, flooding
of ICMP messages start by RSU.
In Figure 20.9, nodes begin monitoring its adjacent nodes, when receive
ICMP messages.
In Figure 20.10, from monitoring, malicious nodes are detected and
isolated from the network. After isolation of malicious nodes, communica­
tion is being continued in the network and malicious nodes are unable to
contact to the RSUs.
Strength-Based Novel Technique for Malicious Nodes Isolation 297

FIGURE 20.6 Attack triggered.

FIGURE 20.7 Getting signal strength information.

20.6 RESULT ANALYSIS

Throughput can be defined as the number of packet data received per unit
time and it is represented in bps.
298 Advances in Data Science and Computing Technology

FIGURE 20.8 Flooding of ICMP messages.

FIGURE 20.9 The monitoring of adjacent nodes.

T = RP / TT (1)
where; T is the throughput; RP is the received packets; TT is the data trans­
mission time (bps).
Strength-Based Novel Technique for Malicious Nodes Isolation 299

FIGURE 20.10 Detection and isolation of malicious nodes.

E2E (end-to-end) delay defined as the time taken between sending of a

packet and it’s receiving on the destination.
D = RT – ST (2)
AD = TD / Number of packets (3)
where; D is the delay; RT is the receiving time; ST is the sending time; AD
is the average delay; TD is the total delay.
Packet loss in a communication is the difference between the generated
and received packets.
PL = GP – RP (4)
where; PL is the packet loss; GP is the generated packets; RP is the received
packets.
Routing Overhead is the number of routing packets required for network
communication.
Routing overhead = Rouring packet counts (5)

Figure 20.11 exhibits, the throughput of the proposed and existing tech­
nique comparison and analyzed that the network throughput is increased at
the steady rate after the malicious node isolation.
300 Advances in Data Science and Computing Technology

Proposed Technique
Existing technique

FIGURE 20.11 Throughput comparison.

Figure 20.12 exhibits, the delay of the proposed and existing algorithm
comparison and analyzed that network delay is reduced when the attack gets
isolated from the network.

FIGURE 20.12 End to end delay comparison.

Figure 20.13 describe the packet loss of the proposed and existing tech­
nique comparison and analyzed that the network packet loss is reduced when
Sybil attack gets isolated from the network. If the packet loss is less then
number of packets drops also less.
Strength-Based Novel Technique for Malicious Nodes Isolation 301

FIGURE 20.13 Packet loss comparison.

Figure 20.14 shows, the routing overhead are the parameter which
measures the extra packets that transmitted in network. The routing overhead
in network is reduced on detection and isolation of attack from the network.
It means fuel emission decreased (Table 20.1).

FIGURE 20.14 Routing overhead comparison.

302 Advances in Data Science and Computing Technology

TABLE 20.1 Performance Parameters Comparison between Existing and Proposed Technique
Time Existing Technique (In respect of Proposed Technique (In respect of
(seconds) number of packets) number of packets)
Through­ E2E Packet Routing Through- E2E Packet Routing
put Delay Loss Overhead put Delay Loss Overhead
4.5 20 46 45 30 25 6 16 16
6 35 70 70 32 55 25 45 17
7 47 75 85 34 90 58 65 18

20.7 CONCLUSION

In this study, it was observed that malicious nodes often enter the network
due to its decentralized existence of network, which really is responsible for
causing different forms of active and passive attacks. This analysis provides
a justification for the identification of malicious nodes that responsible for
triggering a Sybil attack on the network. The simulation of implemented
technique has been done on NS2 and the result shows that overall performance
increased of the network. Future work will include a further improvement
of our technique by removing the chances of information spoofing in the
network.

KEYWORDS

• broadcasting
• decentralized
• path establishment
• self-configuring
• Sybil

REFERENCES

Ahmad, A., Doughan, M., Gauthier, V., Mougharbel, I., & Marot, M., (2010). Hybrid multi-
channel multi-hop MAC in VANETs. In: Proceedings of the 8th International Conference
on Advances in Mobile Computing and Multimedia (pp. 353–357).
Strength-Based Novel Technique for Malicious Nodes Isolation 303

Balamaha, L. D., & Shankar, M. K. V., (2014). Sybil attack detection with reduced bandwidth
overhead in urban vehicular networks. International Journal of Engine ring Trends and
Technology (IJETT), 12, 578–584.
Camp, T., Boleng, J., & Davies, V., (2002). A survey of mobility models for ad hoc network
research. Wireless Communications and Mobile Computing, 2(5), 483–502.
Chang, S., Qi, Y., Zhu, H., Zhao, J., & Shen, X., (2011). Footprint: Detecting Sybil attacks in urban
vehicular networks. IEEE Transactions on Parallel and Distributed Systems, 23(6), 1103–1114.
Chauhan, P., & Mandoria, H. L., (2017). An empirical study of vehicles communication in
vehicular ad-hoc network. In: International Journal of Innovative Research in Science,
Engineering and Technology (Vol. 6, pp. 3763–3769).
Haas, J. J., Hu, Y. C., & Laberteaux, K. P., (2009). Real-world VANET security protocol
performance. In: GLOBECOM 2009–2009 IEEE Global Telecommunications Conference
(pp. 1–7). IEEE.
Hao, Y., Tang, J., & Cheng, Y., (2011). Cooperative sybil attack detection for position-based
applications in privacy preserved VANETs. In: 2011 IEEE Global Telecommunications
Conference- GLOBECOM 2011 (pp. 1–5). IEEE.
Iqbal, S., Chowdhury, S. R., Hyder, C. S., Vasilakos, A. V., & Wang, C. X., (2009). Vehicular
communication: Protocol design, testbed implementation and performance analysis. In:
Proceedings of the 2009 International Conference on Wireless Communications and Mobile
Computing: Connecting the World Wirelessly (pp. 410–415).
Kwag, S. J., & Lee, S. S., (2006). Performance evaluation of IEEE 802.11 ad-hoc network in
vehicle to vehicle communication. In: Proceedings of the 3rd International Conference on
Mobile Technology, Applications and Systems (pp. 47-es).
Malla, A. M., & Sahu, R. K., (2013). Security attacks with an effective solution for dos attacks
in VANET. International Journal of Computer Applications, 66(22).
Mustafa, B., & Raja, U. W., (2010). Issues of Routing in VANET. School of computing at
Blekinge Institute of Technology.
Peel, G., Michielen, M., & Parker, G., (2001). Some aspects of road sweeping vehicle
automation. In: 2001 IEEE/ASME International Conference on Advanced Intelligent
Mechatronics. Proceedings (Cat. No. 01TH8556) (Vol. 1, pp. 337–342). IEEE.
Rabieh, K. M., & Azer, M. A., (2011). Combating sybil attacks in vehicular ad hoc networks. In:
Recent Trends in Wireless and Mobile Networks (pp. 65–72). Springer, Berlin, Heidelberg.
Rajamani, R., (2000). On spacing policies for highway vehicle automation. In: American
Control Conference. Chicago, Illinois.
Rawat, A., Sharma, S., & Sushil, R., (2012). VANET: Security attacks and its possible
solutions. Journal of Information and Operations Management, 3(1), 301.
Raya, M., & Hubaux, J. P., (2007). Securing vehicular ad hoc networks. Journal of Computer
Security, 15(1), 39–68.
Uchhula, V., & Bhatt, B., (2010). Comparison of different ant colony-based routing algorithms.
IJCA Special Issue on MANETs, (2), 97–101.
Wu, H., Palekar, M., Fujimoto, R., Guensler, R., Hunter, M., Lee, J., & Ko, J., (2005). An
empirical study of short-range communications for vehicles. In: Proceedings of the 2nd
ACM International Workshop on Vehicular ad Hoc Networks (pp. 83, 84).
Xiao, B., Yu, B., & Gao, C., (2006). Detection and localization of sybil nodes in VANETs. In:
Proceedings of the 2006 Workshop on Dependability Issues in Wireless Ad Hoc Networks
and Sensor Networks (pp. 1–8).
CHAPTER 21

SECURING THE INFORMATION USING

COMBINED METHOD
BIJOY KUMAR MANDAL, SAPTARSHI ROYCHOWDHURY,
PAYEL MAJUMDER, and ANWESA DAS
NSHM Knowledge Campus, Durgapurs, West Bengal, India,
E-mail: [email protected] (B. K. Mandal)

ABSTRACT

In this project, we propose an algorithm to hide data inside image using

image steganography technique. Using image steganography, image can be
secured that has to be sent over the internet. There are two images one is
cover image and another one is secret image with a secret message that need
to be transferred over the network. The whole process is divided into two
parts first part is encryption that deals with the process of converting plain
text into cipher text using AES algorithm and the second part deals with
decryption using DES algorithm to get the original plain text from a cipher
text.

21.1 INTRODUCTION

With the help of image processing, we have proposed an algorithm by

combing the two method, cryptography, and steganography that is used to
provide security to the secret message by hiding the way of communication
(Fridrich, Goljan, and Soukal, 2014). Steganography is the process of hiding
the relevant information over the network inside an image (Petitcolas,
Anderson, and Kuhn, 2008).
306 Advances in Data Science and Computing Technology

Nowadays, technology is growing very fast. Everything is dependent on

the internet and hence security is must more needed (Cheddad et al., 2010).
So, steganography is used to make it more secure to transform the informa­
tion over the network in more way that is convenient. Here we will use AES
algorithm (which convert the plain text into the cipher text) to make It more
secure and only the authorized user can access (Cheddad et al., 2009). Those
data which made it safe from the cyber-crime and on another algorithm
which is DES algorithm (which is reverse process of AES algorithm means
extract the plain text from cipher text). Steganography is basically based
on message hiding and image stitching which provide double security and
intruders cannot access the useful information (Liu and Tsai, 2007). In the
growing age of technology, where technology is used by a common man in
day to day life, where each second millions of messages are transmitting
across the world wide over network. Where the world is increasing with
technology, the cyber-crime is also increasing, so providing the network
security is not sufficient. We need extra security while transferring messages
or images. Using image steganography, we proposed an algorithm to provide
security to the image, which is being transferred rapidly in day-to-day life
(Girling, 1987).

21.1.1 IMAGE PROCESSING

Image processing is a process of converting an image into digital form and

some operation can be performed on it, to extract some useful information from
the image. In the process of image, processing input is image say photograph,
any video frame and output can be image, or it may be characteristic of that
image (Manfred, 1989). In this age of technology, everything is performed
digitally so image processing play a great role on it in many aspects. In
Image processing system, images are treated as two-dimensional signals.
Image processing includes this mention below steps:
• Import of an image through via image acquisition;
• Recognizing an image;
• The result can be altering image or report based on analysis of image.
Major reason why this encryption and decryption is used is security while
we transferring the message or media through internet or other communica­
tion network (Wendzel and Keller, 2016). As lots of message and information
is transformed from one place to another through internet media, which need
security from hackers or unauthorized users. Proving this security.
Securing the Information Using Combined Method 307

21.1.2 ENCRYPTION

The process where the plain text is transformed into unreadable format. This
is done by using an algorithm for unauthorized users. Here the message or
the data is encoded for which unauthorized user cannot access. The types of
encryption are:
1. Symmetric Encryption: In this encryption, the key that is used for
encryption and decryption is same. Therefore, this use the same key
for the message need to send.
2. Public Key Encryption: In this, the algorithm use both the keys,
i.e., one for public that is called public key and other for authorized
user, which is called secret key. In addition, these two keys are linked
with each other.

21.1.3 DECRYPTION

Decryption is a process of converting the encrypted data into its original

form. It is opposite process of encryption. Only authorized user can decode
this encrypted data because the help of secret key. Only can do decryption
those user can decrypt the information that have a valid key for that encrypted
information (Krzysztof, 2010; Rivest, Shamir, and Adleman, 1978).

21.2 STEGANOGRAPHY

Steganography has vast effect on nowadays world and introduced few new
technologies. It also provides the security during messaging in mobile. This
Steganography word came from two Greek words, one is steganos and
other is graptos. The first steganos means Covered whereas graptos means
writing, which makes the meaning “cover writing.” Steganography means
hiding of message in specific medium like audio, video, and many more.
It uses multimedia as cover image as we know in steganography, we use
one message which in kept inside an image (known as cover image) and
the encrypted image is called stego image (Junod, 2001). People send lots
of private and legal message through internet or the other communication
network and security is key term for them, for those purpose; steganography
is used where actual purpose is to protect the content or the information from
the attack of unauthorized user (hacker) without altering the actual message.
308 Advances in Data Science and Computing Technology

Cryptography and steganography are two terms where cryptography is

use to protect the data and steganography is use for hiding the data/message/
information. There are different types of steganography:
1. Image Steganography: It is used to hide secret message inside an
image. It uses LSB algorithm to hide the message.
2. Audio Steganography: It is used to hide secret message in the audio
using digital representation.
3. Video Steganography: It is combination of image steganography and
audio steganography. Video Steganography is to hide large amount of
data.
4. Text Steganography: It is most difficult steganography technique
because texts have less redundancy compare to audio or image.

21.3 CRYPTOGRAPHY

Cryptography is an important term that is used to provide network security.

It is mainly associated with the converting the plain text into non-readable
text and vice-versa. Cryptography is used for data security as well as for
user authentication while cryptanalysis is the process where cryptosystem is
break. Either the algorithm is alike to each other or you can say two sides of
coin. Steganography does not alter the structure of original message whereas
in cryptography it affects the original secret message (Bijoy, Debnath, and
Xiao-Zhi, 2019). In steganography, the communication of message is hidden
while in cryptography, encryption method is done (Figure 21.1).

FIGURE 21.1 Process of cryptography.

21.3.1 SYMMETRIC KEY CRYPTOGRAPHY

Symmetric key cryptography is also known as private key cryptography. In

this cryptography, a single key is used for both encryption of plain text and
decryption of cipher text. The process of key exchange (means the key itself
is exchanged between sender and receiver) constitutes a problem. In simple
words if the key is revealed then beyond the sender and receiver than there is
a chance to decrypt the message.
Securing the Information Using Combined Method 309

21.3.2 ASYMMETRIC KEY CRYPTOGRAPHY

Asymmetric key cryptography is also known as public key cryptography. In

this cryptography, both public as well as private key are used to decrypt the
data. Here public key is known by everyone and the private key is kept as
secret. Either of the key can be used to encrypt the message, the opposite key
is used for decrypt the message.

21.3.3 HASHING

Hashing is the method of cryptography where one form of data is converted

into a unique string of text. Hashing can be done with any size or type.

21.4 ALGORITHM

1. Input Image: Here the secret message is hidden in selected file. The
message can be in text form also.
2. Image Quality Selection: The quality of output image is selected.
The more quality of image the less data can be hidden.
3. AES Encryption: The secret image is encrypted through the AES
Algorithm. It provides the security id by mistake if unauthorized
user decodes the image file it can’t access or encrypt the information
encrypted by AES.
4. LSB Algorithm: LSB can hide a huge amount of information to the
lower bits of data/information.
5. Encoded and Encrypted Image: Here, the message which is hidden
in image is ready to send to the user. But the user must have that
secret key to decrypt or to open the algorithm (AES).

21.4.1 ON SENDER SIDE

 Step 1: At first, take the inputs the message and any image as image1.
and another image as image2.
 Step 2: Message and image1 need to be broken into n number of
equal parts.
 Step 3: Message or information are encrypted using AES (keys 512)
Algorithm.
310 Advances in Data Science and Computing Technology

 Step 4: Do random dynamic mapping between the sub-parts of image1

and the sub-part of encrypted message.
 Step 5: After the mapping, sub-part of the encrypted message is
hidden with the sub-part of the image1 using LSB algorithm.
 Step 6: The hidden information is again encrypted using AES
algorithm.
 Step 7: Now, all sub-part are hide into the image 2 using LSB algo­
rithm. After that all the sub-part hide into image 2.
 Step 8: Again, image2 is broken into n number of equal parts.
 Step 9: Each part then separately sent to the recipient.

21.4.2 ON RECEIVER SIDE

 Step 1: On the receiver’s end, the sub parts are received.

 Step 2: Now, the parts image need to be merged based on their index
values to regain original cover image.
 Step 3: Process of decryption of secret image is done using AES
algorithm same as encryption, but in reverse order.
 Step 4: Now, get original message using LSB algorithm.
 Step 5: Apply AES algorithm for decryption.
 Step 6: Message and image need to be merged.
 Step 7: We obtain the secret image on receiver end.

21.4.3 LEAST SIGNIFICANT BIT (LSB)

1. Embedding of Text in Image:

i. Calculating the pixel (IMAGE) need to encrypt;
ii. Make a loop through the pixels of the image to be encrypted;
iii. In each passing round we get the RGB Value of pixel image;
iv. Make each RGB value to zero;
v. Get or extract the character from the hidden binary form and
hide the 8 bit binary code in pixel (LSB) of image.
2. Extracting the Embed Message from the Image:
i. Calculate the pixel;
ii. Loop till one finds 8 consecutive zeros in pixel of image;
iii. Convert the picked LSB from pixel element to character.
Securing the Information Using Combined Method 311

21.5 RESULTS

The results are shown in Figures 21.2–21.7.

FIGURE 21.2 Secrete image for input.

FIGURE 21.3 Cover image for input.

312 Advances in Data Science and Computing Technology

Sender side.
FIGURE 21.4
Securing the Information Using Combined Method 313

Encrypted image.
FIGURE 21.5
314 Advances in Data Science and Computing Technology

Sender side out after encryption process.

FIGURE 21.6
 Securing the Information Using Combined Method

FIGURE 21.7 Receiver side output after final process.

315
316 Advances in Data Science and Computing Technology

21.6 CONCLUSION

Nowadays, security is major aspect so we have done this project. In this project
we used steganography technique which is called image steganography. The
data is hidden inside the image which is called stego image. The main aim of
this project to enhance security. In future, to provide greater security to the
data we can use different keys for encryption and decryption. Audio can also
embed in carrier media. Size of images and message can be different. It will
provide an effective way to select an image to hide a message. A secret key is
needed to decode the secret message. Steganography has much use in future.
This is double checker security for transfer of message through internet. Its
application is in use of protection of data alteration and storing confidential
data. Accessing of digital media/data that is stored for authorized user can
also be control by steganography.

KEYWORDS

• AES
• cipher text
• data encryption standard (DES)
• LSB
• steganography

REFERENCES

Bijoy, K. M., Debnath, B., & Xiao-Zhi, G., (2019). “Data security using 512 bits symmetric
key based cryptography in cloud computing system.” In: International Journal of Advanced
Science and Technology (Vol. 129, pp. 1–10).
Cheddad, A., Condell, J., Curran, K., & Mc Kevitt, P., (2009). “A skin tone detection algorithm
for an adaptive approach to steganography.” In: Signal Processing (Vol. 89, No. 12, pp.
2465–2478).
Cheddad, A., Condell, J., Curran, K., &. Mc Kevitt, P., (2010). “Digital image steganography:
Survey and analyses of current methods.” In: Signal Processing (Vol. 90, No. 3, pp. 727–752).
Fridrich, J., Goljan, M., & Soukal, D., (2014). “Searching for the stego key.” Proc. SPIE,
Electronic Imaging, Security, Steganography, and Watermarking of Multimedia Contents,
70–82.
Securing the Information Using Combined Method 317

Girling, C. G., (1987). “Covert channels in LAN’s.” IEEE Transactions on Software Engineering,
13(2), 292–296.
Junod, P., (2001). “On the complexity of Matsui’s attack.” Selected Areas in Cryptography
(Vol. 2259, pp. 199–211). Lecture Notes in Computer Science.
Krzysztof, S., (2010). “HICCUPS: Hidden communication system for corrupted networks.”
In: Proc. of the Tenth International Multi-Conference on Advanced Computer Systems
ACS’2003 (pp. 31–40).
Liu, T. Y., & Tsai, W. H., (2007). “A new steganographic method for data hiding in Microsoft
word documents by a change tracking technique.” In: IEEE Transactions on Information
Forensics and Security (Vol. 2, No. 1, pp. 24–30).
Manfred, W., (1989). “Covert channels in LAN protocols.” Local area network security.
Lecture Notes in Computer Science, 396, 89–101.
Petitcolas, F. A. P., Anderson, R. J., & Kuhn, M. G., (2008). “Information hiding: A survey.”
Proceedings of the IEEE (special Issue), 87(7), 1062–1078.
Rivest, R. L., Shamir, A., & Adleman, L., (1978). “A method for obtaining digital signatures
and public-key cryptosystems.” Communications of the ACM, 21(2), 120–126.
Wendzel, S., & Keller, J., (2016). Low-attention forwarding for mobile network covert
channels. 12th Joint IFIP TC6 and TC11 Conference on Communications and Multimedia
Security (CMS) (Vol. 7025, pp. 122–133). Lecture Notes in Computer Science.
CHAPTER 22

IMPLEMENTATION OF DNA
CRYPTOGRAPHY IN IoT USING
CHINESE REMAINDER THEOREM,
ARITHMETIC ENCODING, AND
ASYMMETRIC KEY CRYPTOGRAPHY
ANANYA SATPATI (DAS), SOUMYA PAUL, and PAYEL MAJUMDER
Department of Computer Science and Engineering, NSHM Knowledge
Campus, Durgapur, Arrah Shibtala Via Muchipara, Durgapur–713212,
West Bengal, India, E-mail: [email protected] (A. Satpati)

ABSTRACT

In today’s world security is an important issue where internet plays an

important role. Network security helps us by protecting our valuable data
from unauthorized access, harmful cryptanalyst, etc. It also confirms that
shared data is kept secure. Stack of protection layer of network security
help us to create a complex cipher. Generally, an intruder focuses on the
file, when it is available in a transmission medium, instead of before
transmitting the file, when it only available to the sending device (like IoT
device). But if we simultaneously increase the security of sending/receiving
devices (like IoT devices) to create more complex cipher with the security
of the transmission process, then our system will be more strong and, in this
chapter, emphasize on this matter. In this chapter, we use a robust security
system that can help any IoT device to create a complex cipher data. In
this chapter encryption process done by knapsack algorithm and public key,
320 Advances in Data Science and Computing Technology

where public key is generated by Chinese remainder theorem. Our system

is a public key crypt system, where spread is generally slower than a private
key crypto system. But here, we can overcome this problem. For more
confusing encrypted data creation we have taken the help of DNA encoding
and Rail Fence technique. After that to compress it, we have used arithmetic
encoding technique.

22.1 INTRODUCTION

In 21st century advancement of technology, storage of big data, fast transmission

of huge data through network, embedded system all are the example of a new
thing known is called Internet of Thinking (IoT) (Zeadally, Das, and Sklavos,
n.d.). The internet of things (IoT) is established on smart and self-conviction
nodes contained in a dynamic and global environment (Raj, Bashar, and
Ramson, 2020). It is one of the most developing technologies, sanctioning
the things to interact world. Recent research has considered IoT indicates
smart and interconnected things capable of sharing their perceptions through
the Internet. Deoxyribonucleic acid (DNA) encoding is a very popular topic
in cryptography domain based research area. DNA is actually a biological
molecules, which having the capacity to store, process, and transmit
information, these capacities are the source of idea of DNA cryptography
(Review on DNA Cryptography, 2019). We generally take the chemical
characteristics of biological DNA sequences in traditional cryptography
system to hide the structure of plain text from an intruder, or in other word
we can say to create more confusion on plain text. DNA computing is the
heart of this innovative method. The methodologies of DNA cryptography
are not coded mathematically, thus, to crack a DNA cryptosystem is very
difficult (Review on DNA Cryptography, 2019). Based on the techniques of
encryption and decryption we can classify cryptosystem in two ways, there
are Symmetric Key Encryption and Asymmetric Key Encryption (Review
on DNA Cryptography, 2019). We all are known Symmetric key encryption
is faster than Asymmetric key encryption, but if we compare based key
transmission process, then we will see strength of security level of asymmetric
key encryption is much more higher than symmetric key encryption, because
every time security is our first priority in a data transmission process. Hence
in this chapter we use asymmetric key DNA encryption technique. In this
chapter the modulo operation of Chinese Remainder Theorem use for key
generation purpose.
Implementation of DNA Cryptography in IoT Using Chinese 321

22.2 PRELIMINARIES

22.2.1 CHINESE REMAINDER THEOREM (FOROUZAN AND

MUKHOPADHYAY, 2003; JIA ET AL., 2019)

In our cryptography system in many cases, we use the concept congruence to

map Z to Zn, it not a one to one mapping, where infinite number of Z map to
one member of Zn. As example we can say 2 mod 10 = 2, 12 mod 10 = 2, 22
mod 2 = 2, and so on. Hence according to modulo arithmetic we can say 2, 12
and 22 are called congruent mod 10. But in Chinese remainder theorem among
all congruent equation variable is only one but it works on different modulo
(Forouzan and Mukhopadhyay, 2003). According to Chinese remainder
theorem we can write, (Jia et al., 2019) if p1, p2. pt are pairwise co-prime posi­
tive integer, then the system of congruence.
x ≡ S1 (mod p1)
x ≡ S2 (mod p2)
x ≡ St (mod pk)
has a unique solution:
t
P
x≡∑ . Si. yi mod P,
i=1 pi

t P

where; P = ∏ i=1 pi , . Si . yi mod P, and “≡” denotes congruence (Jia et al.,

pi
2019).

22.2.2 DNA ENCODING (LUE ET AL., 2007; BISWAS ET AL., 2019;

PELLETIER AND WEIMERSKIRCH, 2002; PAUL ET AL., 2017;
NASKAR ET AL., 2019)

The concept DNA sequence comes from biological. By using different

characteristics of biological DNA like capacity to store, process, and
transmit information, etc., in cryptography field, we can relate these two
different fields. Characteristics of DNA cryptography is the merging form
of biological and computational characteristics, which plays an important
role to increase the security level in a cryptography process (Lue et al.,
2007). Strength of security level of DNA cryptography is much higher than
traditional cryptosystems, hence it is very popular in modern cryptosystem
(Biswas et al., 2019; Pelletier and Weimerskirch, 2002). A DNA sequence is
322 Advances in Data Science and Computing Technology

a combination of four nucleic acids: A (adenine), T (thymine), C (cytosine)

and G (guanine). Depend on base pairing rules, the purine-adenine (A) and
the pyrimidine-thymine (T) create one pair, and the pyrimidine-cytosine (C)
and the purine-guanine (G) create another pair. After following various char­
acteristics of different nucleic acids, we can say A and T are complementary,
as well as the characteristics of G and C are also complemented, where {00,
01, 10, and 11} are the representation of {A, T, G, C} respectively (Biswas
et al., 2019). Simple DNA structure shown in FIGURE 1 and corresponding
all values of four nucleic acids base on encryption and decryption map rules
for DNA sequence shown in Table 1.

FIGURE 22.1 Simple DNA structure (Paul et al., 2017; Naskar et al., 2019).

TABLE 22.1 Encryption and Decryption Map Rules for DNA Sequences (Paul et al., 2017;
Naskar et al., 2019)
A T C G
Rule 1 00 11 10 01
Rule 2 00 11 01 10
Rule 3 11 00 10 01
Rule 4 11 00 01 10
Rule 5 10 01 00 11
Rule 6 01 10 00 11
Rule 7 10 01 11 00
Rule 8 01 10 11 00

22.2.3 XOR ALGEBRAIC OPERATION FOR DNA SEQUENCE

XOR operation of DNA sequence is very useful in encryption process of

cryptography. XOR operation for DNA sequences means bit wise XOR
Implementation of DNA Cryptography in IoT Using Chinese 323

operation between corresponding binary values of nucleic acids of DNA

sequence. Resultant value of XOR operation between various DNA compo­
nents, which are used in the encryption system is shown in Table 22.2.

TABLE 22.2 XOR Operation for DNA Sequences (Paul et al., 2017; Naskar et al., 2019)
⊕ A T C G
A A T C G
T T A G C
C C G A T
G G C T A

22.2.4 ASYMMETRIC KEY CRYPTOSYSTEM (KAHATE, 2003;

ASYMMETRIC CRYPTOGRAPHY), N.D.)

In Asymmetric key cryptography two different keys are used, one for
encryption another for decryption (Kahate, 2003). There are public key
and private key, hence asymmetric key cryptosystem also called public key
cryptography. The general idea of asymmetric key cryptosystem shown in
FIGURE 2. The recipient public key use for encryption purpose and recipient
private key use for decryption purpose.
 Advantages [Asymmetric Cryptography (Public Key Cryptog­
raphy), n.d.]:
• In asymmetric key no key exchange is required, no key distribution
problem arise here.
• Strength of security is higher than symmetric, as private key never
disclose among to all.
• In asymmetric key cryptography sender has no scope to gainsay
regarding sending message, as non-repudiation characteristics
supported by asymmetric key cryptography.
 Disadvantages [Asymmetric Cryptography (Public Key Cryptog­
raphy), n.d.]:
• In long size massage decryption, generally asymmetric key algo­
rithm is not used, because asymmetric key process is compara­
tively slower than symmetric key algorithm.
• If receiver loses his/her private key then message decryption
never possible.
• If private key of a person is revealed by an unauthorized person,
then unauthorized person can also read all secret message’s owner.
324 Advances in Data Science and Computing Technology

FIGURE 22.2 Asymmetric cryptography.

22.2.5 RAIL FENCE TECHNIQUE (KAHATE, 2003)

Plain text to cipher text transmission have two technique, one is substitution
and another is transposition technique (Kahate, 2003). Rail Fence technique
is one of the popular transposition techniques, it also called “zigzag cipher.”
The Rail Fence technique algorithm (Kahate, 2003) is:
• Write down the plain text as a diagonal order;
• After writing diagonal order, read the plain text in row wise.
The pictorial representation of Rail Fence Technique shown in bellow:

Here “key” value define the number of rows.

Implementation of DNA Cryptography in IoT Using Chinese 325

22.3 PROPOSED SCHEME

In our scheme system takes plan text and encrypt it using asymmetric key
cryptographic scheme. At first in the sender uses a public key of receiver
which is created by Chinese Remainder Theorem and Knapsack algorithm.
Similarly, Knapsack algorithm again use at the last step of decryption. The
Knapsack algorithm is described in bellow:
• We have used key pair of public key and private key in the knapsack
algorithm.
• The public key is used only for encryption process and will be known
to everyone and whereas the private key is used during decryption
process (Asymmetric Cryptography (Public Key Cryptography, n.d.)).
• The elements which are from a predefined set of numbers are in
knapsack are provided, if the sum is Known then we can find out the
sum of the numbers quickly of the knapsack, rather it is not an easy
job to find those elements which has participated in the knapsack to
generate the sum.
Following steps are to followed for knapsack algorithm:
Let N = {N_1, N_2…………, N_n} be a predefined set and
I = (I_1, I_2…………., I_n) be a solution set in which I_i is either 0 or 1.

The elements’ sum in knapsack can be calculated as:

S = knapsack Sum (N, I) = N_1I_1 + N_2I _2+………….…+N_nI_n.

• Objective is to find the solution set (I1, I2,………………., In).

• When knapsack Sum(N,I) is given then it is not difficult to find S but
if S&N is given then finding the value of X is difficult which is the
case of inverse of knapsack Sum(N,I).
 Super-Increasing Sets:
• It is easy to find knapsack Sum and its inverse if the sets S is
super increasing.
• Anon empty set is called super-increasing if every nth term
element >= the sum of all (n–1) elements previous to it.
N −1
i.e., N_i> = ∑ i=1 N _ i
• The algorithm for knapsack and its inverse is given as follows:
326 Advances in Data Science and Computing Technology

 Algorithm:

 Example: Let us consider N = [3, 7, 12, 30, 60, 115] and S = 82. To
find the vector I?
i Ni S S ≥ Ni Ii S,S – Ni × Ii
6 115 82 No 0 S←82 – 115 × 0 = 82
5 60 82 Yes 1 S←82 – 60 × 1 = 22
4 30 22 S 0 S←22 – 30 × 0 = 22
3 12 22 Yes 1 S←22 – 12 × 1 = 10
2 7 10 Yes 1 S←10 – 7 × 1 = 3
1 3 3 Yes 1 S← 3 – 3 × 1 = 0

So, the solution set is I = [1,1,1,0,1,0].

FIGURE 22.3 Merkel-Hellman Knapsack cryptosystem.

Implementation of DNA Cryptography in IoT Using Chinese 327

Consider Vivek and Arun want to communicate with each other.

1. Generation of Key:
i. Let I= I_1, I_2,……….I_n be a super-increasing set of values.
ii. Choose a prime p such that p > I_1 + I_2 +…….+ I_n and a
number q with 1 <q < p.
iii. Arun calculates his public set of items, N_i = q × I_i mod p.
iv. The public key is N and private key is I,q and p.
2. Encryption Process:
i. Vivek encrypts the M using S= knapsack_Sum(M,N) algorithm.
ii. Vivek then sends S as ciphertext to Arun.
3. Decryption Process:
When Arun receives S from Vivek, he does the following:
i. Calculates S’ = q^{–1} × S mod p;
ii. Uses M’ = inverse_knapsackSum(S,’I);
iii. Permutes M’ to find M which is original plaintext.
In our proposed algorithm encryption happens in two phases. In our 1st
phase after getting the plain Text in binary form we encrypt the data using
knapsack algorithm but difference is that we use Chinese remainder theorem
to generate the pair of keys instead of modular transformations (Asymmetric
Cryptography (Public Key Cryptography), n.d.), in our 2nd phase of encryption
we have used DNA encoding technique of cryptography followed by Rail
Fence technique which creates more confusion in data to be send and then
we finally compress the encrypted data using Arithmetic encoding algorithm
then we send this to receiver.
In decryption end we have applied reverse process of encryption.

22.4 RESULT AND DISCUSSION

The design of our proposed cryptographic system is discussed in detail by

using Knapsack followed by DNA cryptography is shown in Figure 22.1.
The recipient broadcast the Knapsack public key to the sender. The sender
encrypts the message using Knapsack algorithm and uses receiver’s knapsack
public key (using Chinese remainder theorem) in the 1st phase of cryptography
and in 2nd phase of cryptography we have used DNA cryptography followed
by rail fence technique then the encrypted data goes through a compression
technique by using arithmetic encoding technique and generates the final
cipher Text and then it is sent to the receiver.
328 Advances in Data Science and Computing Technology

FIGURE 22.4 Process diagram.

After receiver receives the ciphertext it first decompresses the received

data then decrypts the decompressed data in just reverse order of two phases
of cryptography where in decryption process of knapsack it uses first reverse
order of rail fence and DNA cryptography and then it uses Knapsack private
key of recipient generated by using Chinese remainder theorem to reveal the
plain text back.
In our proposed algorithm we tried to test the confidentiality and Integrity
of the original message content. So, in our algorithm one testing is done for
checking the encryption and decryption of messages (data files). And the
second test is for the generation of key of knapsack algorithm. Each test has
conducted has successfully generated plain text to cipher text and cipher text
back to plain text. Our Key generation have also been done successfully by
generating public key and private key encryption and decryption process,
respectively.
In the following table, it can be seen that this test successfully returns the
cipher text to plaintext.
Implementation of DNA Cryptography in IoT Using Chinese 329

Encryption and Decryption of the Message

Encryption Decryption
1. Plaintext: 100101111
Cipher text: 5 18 25

Knapsack Public Key: 1

Knapsack Private Key: {5,7,13}

Ciphertext-5 18 25
Plaintext: 100101111

2. Plaintext: 111100101
Cipher text: 23 5 16

Knapsack Public Key: 36

Knapsack Private Key:{5,7,11}

Ciphertext-23 5 16
Plaintext: 111100101

3. Plaintext: 101111100
Cipher text: 58 69 7

Knapsack Public Key: 1223

Knapsack Private Key:{7,5,51}

Ciphertext-58 69 7
Plaintext: 101111100

22.5 CONCLUSIONS

Following conclusions can be drawn from the analysis and obtained result
of our project is:
• Our proposed algorithm using Knapsack clubbed with DNA cryptog­
raphy will give new class of cryptography where the two phases of
cryptography will contribute more towards the secure transmission of
data and will be less prone to different types of attacks.
330 Advances in Data Science and Computing Technology

• As our algorithm has successfully beat the confidentiality and integrity

parameter of network security, we can comfortably use it in the field
of IoT for security of devices.

KEYWORDS

• arithmetic coding
• asymmetric key
• Chinese remainder theorem
• cryptosystem by using knapsack
• data encryption
• DNA encoding
• rail fence technique

REFERENCES

Biswas, M. R., Alam, K. M. R., Tamura, S., & Morimoto, Y., (2019). A technique for
DNA cryptography based on dynamic mechanisms. Journal of Information Security and
Application, 48, 102363.
Forouzan, B. A., & Mukhopadhyay, D., (2003). Cryptography and Network Security (3rd
edn.). McGraw-Hill; Chennai.
Jia, X., Wang, D., Nie, D., Luo, X., & Sun, J. Z., (2019). A new threshold changeable secret
sharing scheme based on the Chinese remainder theorem. Information Sciences, 473,
13–30.
Kahate, A., (2003). Cryptography and Network Security (1stedn.). McGraw-Hill; New Delhi.
Kate Brush (2020). Asymmetric Cryptography (Public Key Cryptography). [Online] https://
searchsecurity.techtarget.com/definition/asymmetric-cryptography (accessed on 08
December 2021).
Lue, M., Lai, X., Xion, G., & Qin, L., (2007). Symmetric-key cryptosystem with DNA
technology. Sci. China Ser. F., 50(3), 324–333.
Mandrita Mondal, & Kumar S. Ray (2019). Review on DNA Cryptography. [Online] https://
arxiv.org/ftp/arxiv/papers/1904/1904.05528.pdf (accessed on 08 December 2021).
Naskar, P. K., Paul, S., Nandy, D., & Chaudhuri, A., (2019). “DNA encoding and channel
shuffling for secured encryption of audio data.” Multimedia Tools and Applications, 78(17),
25019–25042. Springer Nature.
Paul, S., Dasgupta, P., Naskar, P. K., & Chaudhuri, A., (2017). “Secured image encryption
scheme based on DNA encoding &chaotic map.” Review of Computer Engineering Studies,
IIETA, 4(2), 70–75.
Implementation of DNA Cryptography in IoT Using Chinese 331

Pelletier, O., & Weimerskirch, A., (2002). Algorithmic self-assembly of DNA tiles and its
application to cryptanalysis. In: Proc. of 4th Annual Conf. on Generic and Evolutionary
Computation (pp. 139–146). Morgan Kaufmann Publishers Inc.
Raj, S. J., Bashar, A., & Ramson, S. R. J., (2020). Elliptical curve cryptography based
access control solution for IoT based WSN. International Conference on Innovative Data
Communication Technologies and Application (pp. 742–749). Springer Nature.
Zeadally, S., Das, A. K., & Sklavos, N. (2019). Cryptographic Technologies and Protocol
Standards for Internet of Things, Internet of Things. https://doi.org/10.1016/j.iot.2019.100075.
CHAPTER 23

ROBUST TECHNIQUE TO OVERCOME

THWARTING COMMUNICATION
UNDER NARROWBAND JAMMING
IN EXTREMELY LOW FREQUENCY OF
CDMA-DSSS SYSTEM ON CODING­
BASED MATLAB PLATFORM
TANAJIT MANNA,1 RAHIT BASAK,2 MAINAK DAS,2 and ALOK KOLE3
1
Electronics and Communication Engineering Department,
Pailan College of Management and Technology, Kolkata–700104,
West Bengal, India, E-mail: [email protected]
Computer Science and Engineering Department, Pailan College of
2

Management and Technology, Kolkata–700104, West Bengal, India

3
Electrical Engineering Department, RCC Institute of Information
Technology, Kolkata–700015, West Bengal, India

ABSTRACT

This chapter illustrates an approach for smooth communication to mitigate

the jamming under narrow band in code division multiple access (CDMA)
system in case of direct sequence spread spectrum (DSSS) single user
system. In security sector improvement based communication under
narrowband frequency is highly demandable. An analytical as well as
experimental comparison of power spectral density (PSD) of CDMA under
narrowband frequency with and without jamming condition is illustrated in
334 Advances in Data Science and Computing Technology

the chapter. By variation of signal power and Barker code, an optimized

system is illustrated supported by bit error rate (BER) and signal to noise
ratio (Eb/N) curve. Experimental fluctuation of signal power provides
us low narrowband jamming as well as jamming free PSD. Variations of
Spreading factor supported by Barker Code illustrate the results in favor
of optimized model of the system under narrow band jamming condition.
The abovementioned unique proposal of CDMA-DSSS system designing,
modeling, and performance analysis are implemented using coding based
MATLAB® platform.

23.1 INTRODUCTION

Nowadays, the need of optimized performance in DS-CDMA communication

system transmitter and receiver both are real active interest as the application
areas for the DSSS communication system is increasing day by day because
of the advantages of such system in real time practice world. In the presence
of narrowband jamming environment, the modeling is an approach toward
the goal of improving the performance of a DSSS-CDMA communication
system. In this work it has shown clearly that the performance of a
DSSS-CDMA receiver system in under the narrowband jamming can be
significantly boosted by modifying the power of the signal (Abimoussa and
Landry, 2000; Lichuan and Hongya, 2005; Raju, Ristaniemi, Karhunen, and
Oja, 2002). Variation of spreading factor with specific security or barker
code is illustrated in the optimization method.

23.2 RELATED WORK

Jamming reduction under very low frequency band in DS-CDMA system

is very cursed job. Anti-jamming authorization in CDMA (Abimoussa and
Landry, 2000) was first blueprint in concerned field since the year around
2000. The employment was flourished in the tangible domain by adding
particular subspace jut model (Lichuan and Hongya, 2005). By the arsenal of
uncommitted component analysis (Ristaniemi, Raju, and Karhunen, 2002)
this research job was clutch in CDMA domain sharply. In the meantime,
this jamming notion was offer in FM technology for abdication of particular
band (Lichuan and Hongya, 2003). Most cabalistic work done in DS-CDMA
detection technique under near-far field (Manna and Kole, 2016).
Robust Technique to Overcome Thwarting Communication 335

23.3 SYSTEM DESIGN AND MODELLING

In this method we introduced a MATLAB® programming base modeling

and jamming mitigate technique. To ingredient the problem we account a
spreading factor with specific value. Depends upon the spreading factor
model generate a barker code. For generate the barker code we depend on
following formula and algorithm:
n N − v +1
(C v )
= ∑ j=0 aj + v n aj n (1)
Where, (Cv)n ≤ 1 for all 1 ≤ v ≤ N+1. The above equation based on auto-
correlation function.
For Barker generation following model was account (Figure 23.1).

FIGURE 23.1 Barker code generation.

After creation of barker code, the modulation technique like QPSK

is selected for operation. After that we create jamming data follows by
channel. Depend upon the spreading code we spread the data. Generation
of additive white gaussian noise (AWGN) channel grade up the noise in the
system significantly, as well as the projected system almost removed the
noise very sharply. In the time of detection, the signal de spread with same
chip. In the receiver edge we nabbing the PSD for analyze the jamming
amount.
336 Advances in Data Science and Computing Technology

Depending on Barker Code we design the system on proposed algorithm

(Figure 23.2).

FIGURE 23.2 Algorithm for proposed model.

Robust Technique to Overcome Thwarting Communication 337

To achieve our goal, power spectral density (PSD) and bit error rate
(BER) calculation like below. For PSD we used following formula.
 mean  x( f ) 2  
Sx( f ) = lim   
  (2)
T →∞ T
 
T depict time period.
MATLAB® programming we have to modify the Eqn. (2) like below:
T /2
Sx ( f ) = lim T
1 ∫ ∫ C (t1 )C(t2 )mean ( n(t )1 n(t ) 2 ) e − j 2π f (t2 −t1 ) dt1dt2 (3)
T →∞ T /2

In the Eq. (3), product of mean n(t) and C(t) treated as noise. So, the inter­
mediate PSD treated as noise which makes the both System more secure. In
the above equation C(t) is the deterministic signal and n(t) is Gaussian with
zero mean.
BER define as:
 β max 
Pb = Q  
 2  (4)
 2 Tb 
=β max  ∫ [ p (t ) − q (t ) ] dt  (5)
N
 

Here; Pb accounted as BER. Filter matched pulse described by difference

of p (t) and q(t). Tb is time period.

23.4 RESULTS AND DISCUSSION

For mitigate the jamming we accounted 4 variations of experimental data. All

the stages we consider 13 as spreading factor value. For first part we consider
5dB SIR value. As per Figure 23.3 no jamming data reflected in PSD. If we
analysis the Figure 23.4, in 0 to 2.5 Hz frequency jamming reflected.
In 13 spreading factor value and 5 dB SIR are not enough for removing
narrowband jamming which reflected in Figure 23.5. In Figure 23.5 smooth
line represent the analytical BER and line with bubble represented BER with
jamming data.
In 13 spreading factor value and 5 dB SIR are not enough for removing
narrowband jamming which reflected in Figure 23.5. In Figure 23.5 smooth
line represent the analytical BER and line with bubble represented BER with
jamming data.
338 Advances in Data Science and Computing Technology

FIGURE 23.3 DS-CDMA PSD with no jamming.

FIGURE 23.4 DS-CDMA PSD with jamming.

In second stage signal to interference ratio is 8 dB. In this condition results

reflected in Figures 23.6, 23.7 and 23.8. Figure 23.6 delimitate no jamming
PSD, whereas Figure 23.7 delimitate the jamming concept with PSD. In
Figure 23.8, the jamming value minimized as per BER representation. The
BER partially minimized compared to Figure 23.5.
Figures 23.9–23.11 represent the results with SIR value 13 dB. Figure
23.9 no data jamming is remains same. In Figure 23.10 jamming mitigation
reflected positively. Here also PSD plot swing in between positive and
Robust Technique to Overcome Thwarting Communication 339

negative half express the distortion in the performance. But the distortion
not in practically, it depends on resolution. In Figure 23.11 it is clear that
jamming and no jamming BER is closed to each other.

FIGURE 23.5 Bit error rate with and without jamming.

FIGURE 23.6 DS-CDMA PSD with no jamming.

340 Advances in Data Science and Computing Technology

FIGURE 23.7 DS-CDMA PSD with jamming.

FIGURE 23.8 Bit error rate with and without jamming.

Robust Technique to Overcome Thwarting Communication 341

FIGURE 23.9 DS-CDMA PSD without jamming.

FIGURE 23.10 DS-CDMA PSD with jamming.

342 Advances in Data Science and Computing Technology

FIGURE 23.11 Bit error rate with and without jamming.

Now, the spreading factor value with 13 and SIR value in 20 dB asso­
ciated Figures 23.12, 23.13, and 23.14. Here, Figures 23.12 and 23.13 are
looking like same, i.e., this set of data removed all the jamming. Figure 23.14
described the results strongly in favor of no jamming.

FIGURE 23.12 DS-CDMA PSD without jamming.

Robust Technique to Overcome Thwarting Communication 343

FIGURE 23.13 DS-CDMA PSD with jamming.

FIGURE 23.14 Bit error rate with and without jamming.

23.5 CONCLUSION AND FUTURE DEVELOPMENT

As per Table 23.1, we calculate all the parameters under spreading factor 13.
It are clearly visible that in same spreading factor high SIR gives up almost
344 Advances in Data Science and Computing Technology

jamming free PSD. But in low SIR PSD quality is very good but effected by
narrow band jamming. As per PSD plots it is concluded that distortion is due
to resolution, not is practically.

TABLE 23.1 Parameter Analysis

Serial Performance Analysis Data
No. Spreading SIR Swing Range of PSD Without PSD with BER
Factor (dB) Jamming Portion Jamming Jamming
of PSD
1 13 5 10–1 to 101 good good 10–3
2 13 8 10 to 10
–1 0.7
good good 10–2
3 13 13 10–5 to 10+5 Merely Merely 10–1
distorted distorted
4 13 Below 100 Merely Merely Almost
20 distorted distorted overlapped

KEYWORDS

• BER
• CDMA
• DSSS
• MATLAB®
• PSD

REFERENCES

Abimoussa, R., & Landry, R. J., (2000). Anti-jamming solution to narrowband CDMA interference
problem. In: 2000 Canadian Conference on Electrical and Computer Engineering. Conference
Proceedings (Vol. 2, pp. 1057–1062). Navigating to a New Era (Cat. No.00TH8492).
Fang-Biau, U., Jun-Da, C., & Sheng-Han, C., (2004). Smart antennas for multiuser DS/CDMA
communications in multipath fading channels. Eighth IEEE International Symposium on
Spread Spectrum Techniques and Applications - Program and Book of Abstracts (IEEE Cat.
No.04TH8738) (pp. 400–404).
Lichuan, L., & Hongya, G., (2003). Time-varying AR modeling and subspace projection for
FM jammer suppression in DS/SS-CDMA systems. The Thirty-Seventh Asilomar Conference
on Signals, Systems & Computers, 1, 623–627.
Robust Technique to Overcome Thwarting Communication 345

Lichuan, L., & Hongya, G., (2005). Subspace projection and time- varying AR modeling for anti­
jamming DS-CDMA communications. In: 14th Annual International Conference on Wireless
and Optical Communications, 2005 (p. 104). WOCC.
Lu, M., Cheng, F., Wei, S., & Gang, Q., (2017). Comparison of detection methods for non-
cooperative underwater acoustic DSSS signals. Signal Processing Communications and
Computing (ICSPCC) 2017 IEEE International Conference, 1–5.
Lu, M., Cheng, F., Wei, S., & Gang, Q., (2018). A Real-Time Detection System for Non-Cooper­
ative Communicates Electronic and Automation Control Conference (IMCEC) 2018 2nd IEEE
(p. 1331). Underwater acoustic DSSS signals based on LabVIEW, advanced information
management 1335.
Manna, T., & Kole, A., (2016). Performance analysis of secure DSSS multiuser detection
under near far environment. International Conference on Intelligent Control, Power and
Instrumentation 2016 (IEEE Conference). doi: 10.1109/ICICPI.2016.7859713 IEEE ISBN
No.:978-1-5090- 2636-4. IEEE Xplore.
Raju, K., Ristaniemi, T., Karhunen, J., & Oja, A., (2002). Suppression of bit-pulsed jammer
signals in DS-CDMA array system using independent component analysis. In: 2002 IEEE
International Symposium on Circuits and Systems (Vol. 1). Proceedings (Cat. No.02CH37353).
Rida, T., Saqib, E., Sobia, J., & Saqib, A., (2019). Performance analysis of multi-user polar
coded CDMA system. Applied Sciences and Technology (IBCAST) 2019 16th International
Bhurban. Conference (pp. 1000–1005).
Ristaniemi, T., Raju, K., & Karhunen, J., (2002). Jammer mitigation in DS-CDMA array system
using independent component analysis. In: 2002 IEEE International Conference on Communi­
cations Conference Proceedings (Vol. 1, pp. 232–236). ICC 2002 (Cat. No.02CH37333).
Seyedi, A., & Saulnier, G. J., (2001). A sub-channel selective orthogonal frequency division
multiplexing spread spectrum system. In: 2001 MILCOM Proceedings Communications for
Network - Centric Operations Creating the Information Force (Cat. No.01CH37277) (Vol.
2 pp. 1370–1374).
Sriyananda, M. G. S., Joutsensalo, J., & Hämäläinen, T., (2013). Interference cancellation
schemes for spread spectrum systems with blind principles. In: 2013 IEEE 27th International
Conference on Advanced Information Networking and Applications (AINA) (pp. 1078–1082).
Tsung-Chi, L., Chia-Cheng, H., Cheng-Yuan, C., Guu-Chang, Y., & Wing, C. K., ((2007).
Study of MFSK/FH-CDMA wireless communication systems without symbol-synchronous
assumption. IEEE Sarnoff Symposium, 1–5.
 PART VII

Miscellaneous Topics in Computing

CHAPTER 24

THE UTILITY OF REGIONAL

LANGUAGE FACILITY IN
ACCOUNTING SOFTWARE FOR
RETAIL MARKET OF WEST BENGAL
MADHU AGARWAL AGNIHOTRI and ARKAJYOTI PANDIT
Assistant Professor, Department of Commerce,

St. Xavier’s College (Autonomous), Kolkata, West Bengal, India,

E-mail: [email protected]

ABSTRACT

The role of accounting software in the realm of business practice is immense as

they provide information that forms the primary reference for a stakeholder’s
decision making process and portfolio management. Most of the process in
accounting software as well as the financial statements reproduced by it is
most often in English. However, in a country like India where only 10% of
the total population is well versed with the English language such accounting
process and information become cumbersome. Moreover, with linguistic
diversity present in every 15 to 20 kms of the country there is an absolute
chance of English being not known to all the accounting professionals
and stakeholders. Added to this the Census Report of 2011 confirms that
only 28% of the total Indian population receive their education in English
language which justifies the preference of regional language in accounting
process. In the backdrop of such a scenario the current research chapter tries
to assess the utility of regional language facility in accounting software
through opinion collected from 90 accounting professionals and stakeholders
350 Advances in Data Science and Computing Technology

in West Bengal, India. The research also tries to explore the effect of regional
language facility in accounting software on intra business communication at
managerial level. At the end the study seeks to understand the role played by
multilanguage software in meeting customer satisfaction.

24.1 INTRODUCTION

The Indian retail market is at present valued at a whooping Figure 24.1

trillion US Dollars with 40 million participants directly connected with the
selling and purchasing process (Economic Times, 2017). The accounting
process of such a huge market is bound to be complex and cumbersome not
only owing to its size but the linguistic diversity present in the stakeholders
(Aryan, Alrabei, and Haija, 2016). In such a scenario, the role of accounting
software in fastening and simplifying the accounting process is immensely
important. It is also utterly necessary that the Accounting software and the
financial statements produced thereon must have adequate clarity, worthy to
be understood by all concerned.
In India, the accounting process and outcomes are mostly developed
in English. However, it must be noted that only 10% of the entire Indian
population are fluent with the English language (Census Report of India,
2011). Moreover, the country boosts of having 22 official languages with
linguistic diversities seen in every 15 to 20 kms (Linguistic Diversity of
India, 1969). In the backdrop of such a varied linguistic mélange, processing,
and preparation of accounting information in regional language not only
makes the use of accounting software more convenient for the user but helps
the stakeholders of business have a better knowledge of the accounting
information. Any information in a known vernacular posse more coherence
than the same conveyed in a less known language (Linguistic Diversity of
India, 1969). Accounting information being highly relevant to the market
participants cannot be an exception. Such a milieu calls for understanding
the utility of regional language facility in accounting software.
In any business sector irrespective of the country it pertains to, the stake­
holders are liable to get full-fledged accounting information pertaining to the
business or market they operate in. According to Professor Laithy Aryan, the
quality and clarity of accounting information plays a vital role in investment
decisions, portfolio management and liquidity analysis (Aryan, 2017). In
India as most of the population is better versed in their regional language
over and above English, it is a dire need to interpret the utility of regional
language facility in accounting software from the point of view of all the
The Utility of Regional Language Facility in Accounting 351

users and makers of accounting information. The present research chapter

analyzes the utility of regional language facility in accounting software and
if the presence of the concerned facility helps in meeting customer satisfac­
tion. Alongside, the research assesses the effect of presence of ‘regional
language facility in accounting software’ on intra business communication
at managerial level.

24.2 REVIEW OF LITERATURE

Researches pertaining to the role of accounting software in business practice

have been conducted multiple times. However, studies related to the relevance
of regional language in such accounting software are scanty. Among the
few researches conducted, the one that deserves mention at the foremost is
‘Multilingual Information Management’ by Ximo Granell (2015; Garnell,
n.d.). The study on the importance of Accounting Software in small business
by Walton (2019) highlights both the significance and the limitations of
current day accounting software in the business atmosphere. The report of
Centre for Development of Advanced Computing (2019) on ‘Multilingual
computing and heritage computing’ (2019) clearly defines the role of multi-
language facility in various government sectors including the Rajya Sabha
[the upper house of the Indian Parliament]. Danet and Herring (2007) in their
book have detailed the use of multilingual facility in communication over the
internet. The research of Public Utility Accounting by Farber and McDonnell
(2012) highlights the need for accounting lucid enough to be understood by
the lay man. A mention must be made of the laudable study on the impact of
accounting software for business performance by Wickramasinghe (2017)
which vividly describes the significant position of accounting software in
business performance. These research chapters provide an evidence for the
importance of multilingual facility in business perspectives.

24.3 RESEARCH GAP

Studies in the past have successfully pointed out the utility of accounting
software in the progress of business. The researches have also been ardent
in assessing the significance of accounting information in business decision
making. However, it must be understood that the accounting information
generated by the accounting software must be lucid and should have enough
clarity to be understood by all concerned. Researchers however, have
352 Advances in Data Science and Computing Technology

confirmed that most of the financial information produced by accounting

software is in English, a language practiced fluently by only 10% of the
total Indian population. Moreover, the Indian subcontinent is home to more
than 700 languages with linguistic diversity in every 15 to 20 kms. Under
such a scenario it is natural that accounting information in English may
appear complicated for the non-English speakers of the Indian population.
But that does not mean the respective persons are not entitled to clarified
accounting information. In this scenario, if the accounting information can
be generated in vernacular language, then much of the problem regarding the
lucidity of such information to the users who are not well versed in English
is solved. Though a lot of studies proclaim the utility of accounting software
in Business practice but none of them defines the clarity that financial
information could produce if prepared in regional languages spoken by the
majority of stake holders. This forms the core area for assessing the utility of
regional language facility in accounting software.
Another important aspect lies in the fact that almost 79% of the accounting
professionals had their education mostly in vernacular language. These
professionals are bound to be more convenient in accounting in a language
they have been taught all throughout their academic life. This necessity is
also a reason to determine the utility of the presence of regional language
facility in accounting software.

24.4 RESEARCH OBJECTIVES

• To explore the utility of regional language facility in accounting software;

• To assess the effect of regional language facility in accounting soft­
ware on intra business communication at managerial level;
• To understand the role played by Multilanguage software in meeting
customer satisfaction.

24.5 RESEARCH METHODOLOGY

The research takes into account the primary data collected from a population
of 90 accounting professionals and users of accounting information across
the state of West Bengal. The selection of the respondents was limited to
only those persons who have been using accounting software for at least 6
months. The population was selected using snow ball sampling technique.
From the period of February 2020 to March 2020 the respondents were asked
The Utility of Regional Language Facility in Accounting 353

20 questions mostly through online and verbal interview. The first part of the
questionnaire contained 10 questions about the respondent’s personal infor­
mation [e.g., Educational qualification, medium of education, etc.]. While in
the second part, the respondents were asked to convey their degree of agree­
ability towards a particular question using Likert’s 5-point scale. The second
part had nine questions which conclusively have formed the variables for the
research. At the end to have a better clarity of research the respondents were
also asked about any potential problems resulting from the use of regional
language in accounting and accounting information generation. The opinion
of the respondents was noted down and the descriptive statistics was used for
the analysis. For the convenience of research, the respondents were divided
into three different groups based on their work experience. The mean score
of the responses of each category were considered to determine the degree of
agreeability of the category for a particular question. The respective groups
with the number of respondents in each one of them is shown in Table 24.1.

TABLE 24.1 Categories of Respondents and Number of Respondents in Each Category

Work Experience Number of Respondents
6 months to 5 years 38
Above 5 years to 10 years 27
Above 10 years 25
Total 90

24.6 ANALYSIS

The opinion of the respondent categories for each of the variables was
recorded in a data sheet. The respondent categories which had a mean score
higher than the overall average score signifies a greater degree of agree­
ability towards a particular variable than the rest. The entire findings are
shown in Table 24.2.
The numerical analysis as presented in Table 24.2 is graphically shown
in Figure 24.1.

24.6.1 INFERENCE FROM TABLE 24.2 AND FIGURE 24.1

The Numerical Analysis of the mean score obtained by the respondents

as represented through Table 24.2 and Figure 24.1 vividly proclaims the
 354
TABLE 24.2 Mean Score Obtained by Each Category of Respondents and Overall Mean Score for Each Variable
SL. Variables Work Experience of Respondents Overall Mean

No. 6 Months to 5 5 Years Above 10 Years Score for Each

Years to 10 Years Above Variable

Vr.1 Presence of regional language facility in accounting software 3.69 4.85 4.35 4.30
makes its use more convenient.
Vr.2 Presence of regional language in accounting software fastens the 3.6 4.62 4.4 4.21
accounting process

Advances in Data Science and Computing Technology

Vr.3 Preference of regional language over English language for better 3.43 4.57 4.21 4.07
internal communication
Vr.4 Preference of regional language over English language for better 4.06 4.21 3.95 4.07
external communication
Vr.5 Presence of regional language facility in accounting software 4.08 4.31 4.05 4.15
helps auditor to retrieve classified information
Vr.6 Business documents prepared in regional language have a 4.12 4.45 4.2 4.25
positive impact on business communication with supplier.
Vr.7 Invoices prepared in regional language is more acceptable to 3.33 4.37 4 3.9
customers over invoices prepared in English
Vr.8 Quotations prepared in regional language is more desirable to 4.30 4.37 4.3 4.32
customers over quotations prepared in English
Vr.9 Information about customer reward, payback terms or any 3.75 4.74 4.2 4.23
incentive given in regional language has more coverage of
customers than if it is mentioned in English.
The Utility of Regional Language Facility in Accounting 355

FIGURE 24.1 Mean score obtained by each category of respondents and overall mean score
for each variable.

favorability of every respondent category towards the utility of regional

language facility in accounting software. The mean score pertaining to each
variable is found to be approx. 4 in Likert scale of 1 to 5. This statistic is a
vivid representation of the agreeability for utility of regional language in
accounting software. A deeper insight regarding the opinion of each category
of respondents as mentioned in Table 24.1 is summarized as follows:
1. Inference obtained from respondents having work experience
of 6 months to 5 years: The respondents having work experience
of 6 months to 5 years have opined that most of them believe that
presence of regional language in accounting software makes the
accounting procedure convenient for them. However, the degree
of agreeability of this category of respondents [3.69] regarding this
issue is less as compared to the overall average [4.3] as well as from
356 Advances in Data Science and Computing Technology

the other two categories of respondents. A striking observation lies in

the fact that in respect to most of the variables the respondents having
work experience of 6 months to 5 years have shown less agreeability
towards the variables as compared to other two respondent catego­
ries and the overall average. The exception is only in the case of Vr.4
where this respective category has opined that they strongly believe
that regional language is more preferred over English language for
external communication. The group’s mean score of 4.06 in this vari­
able is just same as the overall mean score of 4.07 of the variable.
Another aspect worthy to be noted is that the same group strongly
believes that preparation of quotations in regional language is more
preferred over the same prepared in English. The mean score of 4.3
which is almost same as the overall mean score [4.32] of this variable
proclaims the same.
2. Inference obtained from respondents having work experience
of above 5 years to 10 years: The category of respondents having
work experience of above 5 years and up to 10 years can be marked
as the group who has most ardently favored the utility of regional
language facility in accounting software. This is evident from the
fact that in the case of all variables the mean score obtained by this
category is more than the overall average score for each variable. In
the case of customer reward and pay back system, this group strongly
emphasize that there can be more coverage of customers if the news
about such rewards and its processing are done in regional language.
The mean score of 4.74 which is almost 0.51 points more than the
overall average proclaims the same.
3. Inference obtained from respondents having work experience of
above 10 years: The most experienced working group of the research
also strongly agrees that there is high utility of regional language in
accounting software. The mean scores of such respondents are in
almost parlance with the overall mean score for each response.

24.7 FINDINGS

To put it in a phrase, the utility of regional language facility in accounting

software is immense as almost all the respondents have strongly agreed to the
viability of the variables. However, a fair amount of diversity in responses is
shown by the respondent categories. To highlight the reason for such diversity
the findings have been categorized for each category of respondents:
The Utility of Regional Language Facility in Accounting 357

1. Findings pertaining to respondents having work experience of

6 months to 5 years: It is evident from the numerical analysis that
the respondents having work experience between 6 months and 5
years agree to a less degree to all the factors relating to the utility
of regional language facility in accounting software in comparison
to the other two categories. The primary reason for such result can
be attributed to the fact that most of the respondents in this category
had been taught in English during their academic life and are well
versed in the language. They feel more convenient in using the
accounting software in English because they are habituated with
accounting terms in English language and find it difficult to prepare
bills and quotations in regional language. In their feedback regarding
the potential problems in using regional language while accounting
they have opined that understanding business documents in regional
language becomes complicated as most of them are not aware of
the accounting terms in vernacular. However, when the question of
external communication and preparation of quotations and invoices
in regional language arise, even this group consider that regional
language is more preferred. This is owing to the fact that even though
the users of the accounting software in this category are comfortable
in English, but the persons studying the accounting statements thus
produced are more comfortable if such statements are reproduced in
regional language.
2. Findings pertaining to respondents having work experience of
above 5 years to 10 years: The respondents having work experience
of above 5 years and up to 10 years strongly agree to the utility of
regional language in accounting software. This category is more
experienced than the previous one and has naturally dealt with
more customers. Most of the respondents of this category had their
medium of education in regional language and are less tech-savvy as
compared to the earlier category. They not only find English language
as complex but the entire process of computerized accounting is
inconvenient for them. The main reason for such a scenario is because
of the fact that they had less access to computer and computing terms
during their school and college life. This category naturally works
faster if regional language facility in accounting software is offered
to them and both internal and external communication in regional
language becomes easier. The group also believes that customers
when explained accounting information in a regional language are
more likely to favor the company and develop reliability towards
358 Advances in Data Science and Computing Technology

the same. This group believes that regional language facility in

accounting software not only increases customer base but also
helps in the company’s brand creation and development by gaining
customer loyalty.
3. Findings pertaining to respondents having work experience of
above 10 years: The most experienced group of respondents also
shows a high affinity towards presence of regional language facility
in accounting software. Their opinion greatly matters as they have
handled more clients than the previous one and have a greater degree
of understanding of the nuances of accounting owing to their work
experience. This group strongly believes the necessity of regional
language facility in accounting software but in no way underestimates
English language. They have handled foreign clients in addition to
domestic clients and believe that both English and regional language
facility in accounting software is necessary as that would make it
regionally viable and internationally acceptable. The reason for such
a perspective is due to the fact that this group of respondents have
become well versed in English language during their tenure of work
as well as possess the knowledge of regional language. The group’s
client and colleague base has more diversity than the previous two
groups. To them business communication and customer satisfaction
can be done both by regional language and English depending on the
sphere of accounting concerned.

24.8 RECOMMENDATION

The role of regional language facility in accounting software is huge as all the
respondents have strongly agreed upon its viability. However, for software to
sustain in the global business scenario, accounting only in regional language
cannot suffice. For operational efficiency presence of regional language
facility is a must but for global expansion English which is the most widely
practiced professional language is required.
The feedback of the respondents indicates that professional accounting
terms [like Balance Sheet, print, etc.], are better understood in English than
in vernacular. Under such a background it is highly desirable that accounting
software have mixed language facility where professional accounting terms are
present in English and the same has been transliterated in regional language.
Owing to the perspective of business expansion on an international basis,
the accounting software demand the presence of both English and regional
The Utility of Regional Language Facility in Accounting 359

language facility simultaneously. Such facility will increase operational

efficiency as well as will provide impetus for global outreach. Accounting
information as produced by the accounting professionals are often found to
be difficult to decipher for the stake holders who does not have much of
accounting knowledge. To tackle the situation, it is an urgent need that the
annual reports and other statements be prepared both in English and regional
language to have more clarified information.
Regional language facility helps in better internal communication
pertaining to break up of accounting process. This need vividly promotes the
utility of regional language facility in accounting software. On a conclusive
note, the research reveals that in order to have a more clarified accounting
process regional language facility is absolute necessity but must be precluded
by the presence of English language in the same.
The entire recommendation can be summarized through the following
model of existing and proposed dimensions of accounting software as shown
in Figure 24.2.

FIGURE 24.2 Model for existing and proposed dimensions of accounting software.

24.9 CONCLUSION

The accounting scenario of India is a massive area with the presence of a

varied level of linguistic diversity. Added to this it must be understood that
the Indian sub-continent is home to 700 languages with linguistic diversity
present in every 15 to 20 kms. For all the stakeholders of the accounting
process interpretation of financial information in the single language of
English is indeed ungainly considering the fact that only 10% of the total
360 Advances in Data Science and Computing Technology

Indian population is fluent with the language. This proclaims the need for
regional language facility in accounting software. However, it must be under­
stood that professional accounting terms are better understood in English
than in regional language. This very reason is worthy enough to convey that
not only regional translation but regional transliteration is equally important
in accounting software.
For any stake holder of accounting process, the financial information is
very relevant and is the best reference for decision making. This rationale
intensely justifies the clarity of accounting information. For a non-English
speaker the clarity of accounting information will be more when such
accounting information is reproduced in vernacular. In the light of such
significant reason the utility of regional language facility in accounting
software is noteworthy. However, it must be kept in mind that in the neces­
sity of the presence of regional language facility in accounting software
the importance of English language in the same cannot be underestimated.
Operational efficiency in a multi linguistic country like India obviously
demands regional language facility but the presence of English alongside
provides the impetus for global expansion. On an ending note, it can be said
that for an accounting software to have utmost efficiency both the presence
of regional language along with English is not only a requirement but it must
form a basic feature of the same.

KEYWORDS

• accounting software
• customer
• linguistic diversity
• multi-language
• regional language
• retail market

REFERENCES

Aryan, L., (2017). The Role of Accounting Information Quality in Management. ResearchGate.
Aryan, L., Alrabei, A., & Haija, A., (2016). The Role of Accounting Information Quality in
Enhancing Cost Accounting Objectives in Jordanian Industrial Companies. ResearchGate.
The Utility of Regional Language Facility in Accounting 361

Census Report of India, (2011). https://censusindia.gov.in/2011-Common/Archive.html.

Danet, B., & Herring, S., (2013). The Multilingual Internet: Language, Culture, and Communi­
cation Online. Oxford University Press, ISBN: 978-0195304794.
Economic Times, (2017). Indian Retail Market Expected to Reach $1 Trillion by 2020. https://
economictimes.indiatimes.com/industry/services/retail/indian-retail-market-expected-to­
reach-1-trillion-by-2020/articleshow/61662729.cms.
Farber, B., & McDonnell, (2012). Public Utility Accounting/ American Public Power Association.
Garnell, X. (2014). Multilingual Information Management (1st edn.). Elsevier, ISBN 97218433
47712.
https://www.cdac.in/index.aspx?id=mlingual_heritage (2019). Centre for Development of
Advanced Computing, Multilingual Computing and Heritage Computing.
Linguistic Diversity of India, (1969). Sodhganga, Chapter III, pp. 92–122.
Walton, A., (2019). The Importance of Accounting Software. Chron.
Wickramasinghe, M., (2017). Impact of accounting software for business performance.
Imperial Journal of interdisciplinary Research, 3(5). ISSN 24541362.
CHAPTER 25

COMPARATIVE ANALYSIS OF STOREY

DRIFT OF A G+9 RC-STRUCTURED
BUILDING DUE TO SEISMIC LOADS IN
VARIOUS SEISMIC REGIONS
BUDHADITYA DUTTA,1 ADITYA NARAYAN CHAKRABORTY,1
AYUSHI SHAH,1 DEBARKA BRAHMA,1 AMIT DEB,2 and
TUSHAR KANTI DE3
1
Undergraduate Student, Department of Civil Engineering,
Amity University Kolkata, Kolkata, West Bengal–700135, India,
E-mail: bdneil3103@gmail (B. Dutta)
Assistant Professor, Department of Civil Engineering,
2

Amity University Kolkata, Kolkata, West Bengal–700135, India

3
Associate Professor and Head of Department, Department of Civil
Engineering, Budge Budge Institute of Technology, Budge Budge,
West Bengal–700137, India

ABSTRACT

Seismic loads are one of the important loads that are to be considered while
designing any structure, like, buildings, power plants, towers, bridges, etc.,
irrespective of using a reinforced-concrete frame or steel frame. The Indian
subcontinent has a history of highly devastating earthquakes over the years.
The main reason behind the high frequency and intensity of earthquakes
is the driving of the Indian plate into the Eurasian plate at a rate of 47
mm/year. As per geographical statistics, about 54% of the landmass of
364 Advances in Data Science and Computing Technology

India is vulnerable to earthquakes. The Indian Standard code for designing

structures against seismic loads is 1893 (Part 1): 2016. As per IS 1893,
India can be divided into 4 seismic zone: Zone II, Zone III, Zone IV, Zone
V, with Zone V experiencing the highest level of seismicity and Zone II the
least. The object of this chapter is to analyze and compare the story drift
of a G+9 RC-structured residential building against seismic loads in the
4 different seismic regions. The software used to carry out the analysis is
ETABS 2016.

25.1 INTRODUCTION

Seismic tremors cause ground movements in an arbitrary design in all direc­

tions having critical flat and vertical ground increasing speeds as function of
time. Designs exposed to ground movements react in a vibratory style. The
most extreme reaction acceleration during the elastic stage relies upon the
normal recurrence of vibration of the design and the scale of the damping.
The most extreme inertia loads following up on a structure during a tremor is
dictated by duplicating the mass by the acceleration (Bulusu, 2016).
The distinctive parameters like intensity, duration, etc., of seismic
ground vibrations depend upon the degree of the earthquake, its depth of
focus, distance from epicenter, properties of soil or medium through which
the seismic waves travel and the soil strata when the structure stands. The
random earthquake motions can be reduced in any three mutually perpen­
dicular directions (Bulusu, 2016). The horizontal direction is normally the
prominent, while vertical acceleration is considered in large-span structures
(Jangid, n.d.). The response of a structure to the ground depends on the
nature of foundation soil, form, material, size, and mode of construction of
structures and the duration and characteristics of ground motion.
The following assumptions are considered while designing earthquake-
resisting structures (IS 1893 (Part 1), 2016):
• Imprudent ground movements of a tremor are complicated, sporadic in
character, changing period and time and of brief length. They may not
cause resonance as pictured under consistent sinusoidal excitations,
besides in tall constructions established on profound delicate soil.
• During the time of the earthquake, any wind, flood, or waves will not
occur.
• For static analysis, elastic modulus of the materials shall be considered,
unless otherwise mentioned.
Comparative Analysis of Storey Drift 365

The components in reinforced concrete casing structures that are

significant under seismic stacking are the joints of sections and bars or floor
slabs, structural walls, spliced areas, shear walls, and coupling beams. All
in all, exceptional consideration ought to be taken for cyclic loadings in any
place where underlying discontinuity exists. Basic regions for such cases
can likewise be the area where there is a chance of plastic hinge formation
(Necdet, 2004).

25.2 DEFINITIONS

As per IS 1893 (Part 1): 2016 (IS 1893 (Part 1), 2016):
1. Design horizontal seismic coefficient, Ah (Clause 6.4.2) (IS 1893
(Part 1), 2016):
Z Sa
*
2 g
Ah =
R

where; Z is the seismic zone factor (IS 1893 (Part 1), 2016); I is the
importance factor (IS 1893 (Part 1), 2016); R is the response reduction
Sa
factor (IS 1893 (Part 1), 2016); is the design acceleration coeffi­
g
cient for different soil types, normalized with peak ground acceleration,
corresponding to natural period T of structure (IS 1893 (Part 1), 2016)
2. Design lateral force, VB (Clause 7.2.1) (IS 1893 (Part 1), 2016)
VB = Ah × W

where; Ah is the design horizontal seismic load (IS 1893 (Part 1),
2016); W is the seismic height of the building (IS 1893 (Part 1), 2016).

25.3 MODEL ANALYSIS

The ETABS 2016 software has been used for 3D model analysis. As per
Figure 25.1 – Seismic Zones of India (IS 1893 (Part 1): 2016), the topography
of India can be divided into four seismic zones (IS 1893 (Part 1), 2016):
1. Zone II: Regions which fall under Low Damage Risk Zone and have
Seismic zone factor, Z = 0.1 (IS 1893 (Part 1), 2016).
366 Advances in Data Science and Computing Technology

2. Zone III: Regions which fall under Moderate Damage Risk Zone
and have Z = 0.16 (IS 1893 (Part 1), 2016).
3. Zone IV: Regions which fall under High Damage Risk Zone and
have Z = 0.24 (IS 1893 (Part 1), 2016).
4. Zone V: Regions which experience highest intensity of earthquakes
and have Z = 0.36 (IS 1893 (Part 1), 2016).
A model G+9 RC-structured building is placed at each of these 4 zones
keeping the material properties and structural members unchanged, unless
needed based on design check results (Table 25.1).

TABLE 25.1 Sites Chosen for Analysis from Different Seismic Zones
Location Seismic Zone Z Soil Type I R
Bengaluru II 0.10 II 1 5
Kolkata III 0.16 II 1 5
Delhi IV 0.24 II 1 5
Guwahati V 0.36 II 1 5

Model characteristics:
• Base of building structure: 20 m × 20 m;
• Number of floors: 10;
• Height of each floor: 3.3 m;
• Height of base floor: 3.5 m;
• Total height of building structure: 33.2 m;
• Aspect ratio: 1.6:1.
Initially, the structure is designed using M25 concrete with 250×400
beams, 450×500 columns and 125 mm thick two-way slabs. Further changes
are made as per requirements, based on the stability of beams and columns
after the design check, so that the building can attain the minimum stable
structure required to resist the seismic loads. The RC designs are made in
accordance with IS 456:2000 (IS 456, 2000).
The loads applied are:
• Dead load (DL) – Wall load, floor finish load, self-load (IS 456, 2000);
• Live Load (LL) (IS 456, 2000);
• Earthquake Loads along X-direction (ELX) and Y-direction (ELY) (IS
456, 2000).
As per assumptions, wind load is neglected.
Comparative Analysis of Storey Drift 367

Soil type (taken from Table 4 – IS 1893 (Part 1): 2016).

• Type I: Rock or hard soil (IS 1893 (Part 1), 2016);
• Type II: Medium soil (IS 1893 (Part 1), 2016);
• Type III: Soft soil (IS 1893 (Part 1), 2016).
Storey drift is defined as the difference between the displacement of
two consecutive storys (Jean-Marie, Nicolas, and Michael, 2010). Storey
displacement is the absolute value of the displacement of the story under
the action of lateral forces (Shariff et al., 2019). Clause 7.11.1.1 of IS 1893
(Part 1): 2016 states that story drift shall not exceed 0.004 times the story
height (Sanyogita, 2019). Therefore, the maximum story drift permissible is
less than (0.004 × 3,300) or 13.2 mm for storys 2 to 10, while, for story 1, it
should be (0.004 × 3,500) or 14 mm.
The rotational reaction of building structures during solid ground move­
ments has ended up being the fundamental driver of fractional or complete
collapse. As of late various experiments have been done to show the seismic
weakness because of building imbalance and mass or firmness abnormality
(Figure 25.2) (Georgoussis, Tsompanos, and Makarios, 2015).

FIGURE 25.1 Plan of model structure

368 Advances in Data Science and Computing Technology

FIGURE 25.2 Elevation of model structure.

Comparative Analysis of Storey Drift 369

25.4 LOAD COMBINATIONS

As per Clause 6.3.1 of IS 1893 (Part 1) 2016, the following load combi­
nations are to be considered for designing a seismic structure (IS 1893
(Part 1), 2016):
• 1.2[DL + IL ± (ELX ± 0.3ELY)] and 1.2[DL + IL ± (ELY ± 0.3ELX)] (IS
1893 (Part 1), 2016);
• 1.5[DL ± (ELX ± 0.3ELY)] and 1.5[DL± (ELY ± 0.3ELX)] (IS 1893
(Part 1), 2016);
• 0.9[DL ± (ELX ± 0.3ELY)] and 0.9[DL± (ELY ± 0.3ELX)] (IS 1893
(Part 1), 2016).

25.5 RESULTS OF THE ANALYSIS

1. Location: Bengaluru (Figures 25.3 and 25.4)

FIGURE 25.3 Storey displacement due to ELX.

370 Advances in Data Science and Computing Technology

FIGURE 25.4 Storey displacement due to ELY.

From the graphs, it is clear that the lateral displacement increases as it

goes from the bottom floor to the top floor. However, the lateral drift starts
decreasing as it goes from the bottom floor to the top floor. We will determine
the story drifts at the storys 10 and 1.

For story10:
i. Due to ELX:
Maximum displacement for story 10 = 16.856 mm

Maximum displacement for story 9 = 16.275 mm

Storey drift = 0.581 mm

ii. Due to ELY:

Maximum displacement for story 10 = 16.970 mm

Maximum displacement for story 9 = 16.381 mm

Storey drift = 0.589 mm

Comparative Analysis of Storey Drift 371

For story 1:
i. Due to ELX:
Maximum displacement for story 1 = 3.051 mm

Maximum displacement for base = 0 mm

Storey drift = 3.051 mm

ii. Due to ELY:

Maximum displacement for story 1 = 3.042 mm

Maximum displacement for base = 0 mm

Storey drift = 3.042 mm

As, the story drifts in both the floors, are less the maximum permissible
drift, for loads acting from both directions, we can consider the structure to
be safe.

2. Location: Kolkata (Figures 25.5 and 25.6)

FIGURE 25.5 Storey displacement due to ELX.

372 Advances in Data Science and Computing Technology

FIGURE 25.6 Storey displacement due to ELY.

For story10:
i. Due to ELX:
Maximum displacement for story 10 = 26.957 mm
Maximum displacement for story 9 = 26.028 mm
Storey drift = 0.929 mm
ii. Due to ELY:
Maximum displacement for story 10 = 27.140 mm
Maximum displacement for story 9 = 26.197 mm
Storey drift = 0.943 mm
For story 1:
i. Due to ELX:
Maximum displacement for story 1 = 4.860 mm
Maximum displacement for base = 0 mm
Storey drift = 4.860 mm
ii. Due to ELY:
Maximum displacement for story 1 = 4.846 mm
Maximum displacement for base = 0 mm
Storey drift = 4.846 mm
Comparative Analysis of Storey Drift 373

As, the story drifts in both the floors, are less the maximum permissible drift,
for loads acting from both directions, we can consider the structure to be safe.

3. Location: Delhi (Figures 25.7 and 25.8)

FIGURE 25.7 Storey displacement due to ELX.

FIGURE 25.8 Storey displacement due to ELY.

374 Advances in Data Science and Computing Technology

For story 10:

i. Due to ELX:
Maximum displacement for story 10 = 40.410 mm

Maximum displacement for story 9 = 39.011 mm

Storey drift = 1.399 mm

ii. Due to ELY:

Maximum displacement for story 10 = 41.261 mm

Maximum displacement for story 9 = 39.815 mm

Storey drift = 1.446 mm

For story 1:
i. Due to ELX:

Maximum displacement for story 1 = 7.149 mm

Maximum displacement for base = 0 mm

Storey drift = 7.149 mm

ii. Due to ELY:

Maximum displacement for story 1 = 7.220 mm

Maximum displacement for base = 0 mm

Storey drift = 7.220 mm

As, the story drifts in both the floors, are less the maximum permissible drift,
for loads acting from both directions, we can consider the structure to be safe.

4. Location: Guwahati (Figures 25.9 and 25.10)

FIGURE 25.9 Storey displacement due to ELX.

Comparative Analysis of Storey Drift 375

FIGURE 25.10 Storey displacement due to ELY.

For story 10:

i. Due to ELX:
Maximum displacement for story 10 = 60.465 mm
Maximum displacement for story 9 = 58.358 mm
Storey drift = 2.107 mm
ii. Due to ELY:
Maximum displacement for story 10 = 60.662 mm
Maximum displacement for story 9 = 58.520 mm
Storey drift = 2.142 mm
For story 1:
i. Due to ELX:
Maximum displacement for story 1 = 10.430 mm
Maximum displacement for base = 0 mm
Storey drift = 10.430 mm
ii. Due to ELY:
Maximum displacement for story 1 = 10.423 mm
Maximum displacement for base = 0 mm
Storey drift = 10.423 mm
376 Advances in Data Science and Computing Technology

As, the story drifts in both the floors, are less the maximum permissible
drift, for loads acting from both directions, we can consider the structure to
be safe.

25.6 DISCUSSION

Using the values of story drift from the analysis, we can plot a graph to
explain the relation of story drift to the seismic zone factor, for both direction
of seismic loads. The seismic zone factor is plotted along X-axis and the
story drift along Y-axis (Figures 25.11 and 25.12).

FIGURE 25.11 Seismic zone factor v/s story drift graph for storys 1 and 10 due to seismic
loads along X-direction.

From the graphs it is observed that the slope for story drift curve for story
1 is much greater than that for story 10. This explains that the inertia forces
acting along the upper part of the structure is greater than that acting along
the lower part. This can cause shearing of the structure, which may result
in concentration of stresses along the walls or joints, leading to failure or
Comparative Analysis of Storey Drift 377

perhaps total collapse. Hence, measures should be taken to reduce the impact
of vibration on the lower floors due to an earthquake.

FIGURE 25.12 Seismic zone factor v/s story drift graph for storys 1 and 10 due to seismic
loads along Y-direction.

Measures to reduce the impact of seismic vibration:

1. Base-Isolation: Base isolation is perhaps the most well-known
methods for securing a construction against quake vibration. It is
an assortment of primary components which ought to significantly
decouple a superstructure from its base, laying on a shaking ground,
accordingly ensuring the underlying uprightness. Base isolation is
quite possibly the most significant assets of tremor designing relating
to the passive underlying vibration control advances. The seclusion
can be obtained by the utilization of different procedures, e.g., rubber
bearings, friction bearings, ball bearings, spring frameworks and
similar different methods. A versatile base separation framework
incorporates a tunable isolator that can change its properties
dependent on the contribution to limit the moved vibration (Behrooz,
Wang, and Gordaninejad, 2014; Seismic Base Isolation, n.d.).
378 Advances in Data Science and Computing Technology

2. Seismic Dampers: Seismic performance of buildings can also

be improved by installing seismic dampers in place of structural
elements, such as diagonal braces. These dampers act like hydraulic
shock absorbers in cars. Dampers absorb a part of the seismic energy
transmitted through them and thus damp the motion of the building.
In case of tall buildings, base-isolation technique cannot be imple­
mented as it can cause the building to over-turn. In such scenarios,
the horizontal displacement can be controlled, which can be done by
using dampers (Vrba, 2020).
Firdous Patel Anjum and L. G. Kalurkar studied the effects of
use of exponential damping mechanism coupled with RC-framed
structure and established responses from various finite element
models and it was observed that exponential dampers improved the
responses (Firdous and Kalurkar, 2018).
3. Shear Walls: These are stiffened walls which have the ability to
resist moment due to lateral loads. They can transfer lateral forces
from roofs and floors to the foundation of a building, thus dissipating
the loads acting on them (Azad and Gani, 2016).
Aainawala et al. did the relative investigation of multistoried R.
C. C. structures with and without shear walls. They applied the quake
burden to a structure for G+12, G+25, G+38 situated in zone II, zone
III, zone IV, and zone V for various instances of shear wall position.
They determined the lateral uprooting and story drift in all the cases.
It was seen that multistoried R. C. C. structures with shear wall is
prudent when contrasted with the one without shear wall. According
to examination, it was concluded that displacement at various levels
in multistoried structure with shear walls is similarly lesser unlike the
R. C. C. building without shear wall (Aainawala and Pajgade, 2014).
4. Moment-Resisting Frames: The column-beam joints in a moment-
resisting frame is capable of handling both shear and moment.
Hence, they eliminate the space limitations of solid shear walls or
braced frames. The column-beam joints are designed, taking into
consideration stiffness, yet allow for some deformation for energy
dissipation, taking advantage of ductility of steel.
5. Energy-Dissipating Devices: Energy-dissipating devices are used
to reduce shaking of the building structure, which occurs when the
building is made more resistive and may damage the vital portions
of the building. Energy dissipates when ductile materials deform in a
controlled way (Necdet, 2004).
Comparative Analysis of Storey Drift 379

Structural and architectural detailing, along with construction

quality control are important factors to guarantee ductility and
natural damping. They also limit the damages to a repairable range.
The prospect of structural and non-structural damage is not avoid­
able without the practical use of energy-dissipating devices.
6. Fiber-Reinforced Concrete: This can be instrumental in reducing
bond degradation within such critical regions of a reinforced
concrete frame. Shear strength of reinforced concrete members can
be increased by the addition of fiber confinement. There is a strong
case for considering fiber reinforced concrete as an alternative to
conventional concrete in reinforced concrete frame structures located
in severe seismic zones (Patnaik and Adhikari, 2012).
An increase of 30% in the punching resistance of slab-column
connection as well as improved ductility of the connection was
recorded by Alexander and Simmonds (Alexander and Simmonds,
1992) due to the addition of corrugated steel fibers.

25.7 CONCLUSION

From the story drift plots, it can be observed that, taking the conditions
of soil type, importance factor and response reduction factor as constant,
for a structure studied at different seismic zones, the drift decreases from
the bottom floors to the top. This is opposite to the trend exhibited by
story displacement. Storey displacement is maximum at the top floor,
decreases towards the lower floors and ultimately becomes null at base or
ground level. Hence, we can conclude that story displacement is directly
proportional to the height of a building, while story drift is inversely
proportional to the height. Storey drift is greater at the bottom as the
bottom floors undergo lesser forces of inertia acting on them, than the
higher floors, due to seismic vibration. Measures should be taken to keep
the impact of seismic loads on the lower floors in check, in order to set
up a stable structure. The various approaches that can be taken to reduce
seismic impact are base-isolation technique, installation of seismic
dampers, setting up shear walls, using moment-resisting frame, energy-
dissipating devices, etc. The story drifts of the storys 1 and 10 of the
model structure do not exceed the maximum drift limit in all 4 seismic
zones, as specified by IS 1893 (Part 1): 2016. Hence the structures can
be considered safe.
380 Advances in Data Science and Computing Technology

The critical bending moment of irregular frames is more than the

customary casing for all structure statures. This is because of abatement in
firmness of building outlines because of setbacks. Subsequently, there is
need for giving greater fortification to irregular frames (Avantika, Vinayak,
and Valsson, 2019).

KEYWORDS

• reinforced concrete frame

• seismic load
• seismicity
• story drift

REFERENCES

Aainawala, M. S., & Pajgade, P. S., (2014). “Design of multistoried R.C.C. buildings with
and without shear walls.” International Journal of Engineering Sciences & Research
Technology, 7(3), 498–510. ISSN: 2277-9655.
Alexander, S. D. B., & Simmonds, S. H., (1992). Punching shear tests of concrete slab-column
joints containing fiber reinforcement. ACI Structural Journal.
Avantika, H. D., Vinayak, D. V., & Valsson, V., (2019). “Effect of stiffness of different shapes
of building to seismic zones.” International Journal of Scientific Research and Engineering
Development, 2(2).
Azad, Md. S., & Gani, S., (2016). “Comparative study of seismic analysis of multistory
buildings with shear walls and bracing systems.” International Journal of Advanced
Structures and Geotechnical Engineering, 5, 72–77.
Behrooz, M., Wang, X., & Gordaninejad, F., (2014). Performance of a new magnetorheological
elastomer isolation system. Smart Materials and Structures, Smart Materials and Structures,
23(4).
Bulusu, U., (2016). “Earthquake Resistant Design of Structures.” https://www.linkedin.com/pulse/
earthquake-resistant-design-structures-udaya-bhaaskar-bulusu-1 (accessed on 08 December
2021).
Firdous, P. A., & Kalurkar, L. G., (2018). “Performance of R.C. frames coupled with
exponential dampers under seismically triggered condition.” International Journal of
Research in Engineering and Technology.
Georgoussis, G., Tsompanos, A., & Makarios, T., (2015). Approximate seismic analysis of
multi-story buildings with mass and stiffness irregularities. Procedia Engineering, 125,
959–966. 10.1016/j.proeng.2015.11.147.
Comparative Analysis of Storey Drift 381

IS 1893 (Part 1), (2016). Criteria for Earthquake Resistant Design of structures – General
Provisions and Buildings (Sixth Revision).
IS 456, (2000). Indian Standard Code of Practice for Plain and Reinforced Concrete (Fourth
Revision).
Jangid, R. S. (2013). Lecture Series on Earthquake Engineering. Department of Civil Engi­
neering, IIT Bombay.
Jean-Marie, R., & Nicolas, I., & Michael, B., (2010). Inelastic seismic analysis of the SPEAR
test building. European Journal of Environmental and Civil Engineering, 14, 855–867.
10.1080/19648189.2010.9693266.
Necdet, T., (2004). “Seismic isolation and energy dissipating systems in earthquake resistant
design.” In: 13th World Conference on Earthquake Engineering. Vancouver, B.C. Canada.
Patnaik, A., & Adhikari, S., (2012). Potential applications of steel fiber reinforced concrete to
improve seismic response of frame structures. NED University Journal of Research.
Sanyogita, S. B., (2019). Seismic analysis of vertical irregularities in buildings. In: Singh, H.,
Garg, P., & Kaur, I., (eds.), Proceedings of the 1st International Conference on Sustainable
Waste Management through Design. ICSWMD 2018 (Vol. 21). Lecture Notes in Civil
Engineering.
Seismic Base Isolation.” (2010). en.wikipedia.org. https://en.wikipedia.org/wiki/Seismic_
base_isolation (accessed on 08 December 2021).
Shariff, M., Owais, M., Rachana, C., Vinu, S., & AsishDubay, B., (2019). “Seismic analysis
of multistorey building with bracings using ETABS.” International Journal of Innovative
Research in Science, Engineering and Technology, 8(5).
Vrba, M., (2020). Hydraulic Shock Absorber in Vehicles and Other Applications [online].
Brno, [cit. 2021-01-29]. http://hdl.handle.net/11012/191756. Bachelor thesis. Technical
University Brno. Faculty of Mechanical Engineering. Department of Automotive and
Transportation Engineering.
CHAPTER 26

A PROPOSAL FOR DISINFECTION

OF CONTAMINATED SUBJECT AREA
THROUGH LIGHT
ANUBRATA MONDAL1 and KAMALIKA GHOSH2
1
Research Scholar, School of Illumination Science Engineering and

Design, (Jadavpur University, Kolkata) and Assistant Professor,

Department of Electrical Engineering, Global Institute of Management

and Technology, West Bengal, India

2
Director, School of Illumination Science, Engineering and Design,

Jadavpur University, Kolkata, West Bengal, India,

E-mail: [email protected]

ABSTRACT

At present world is passing through a pandemic situation due to sudden ingress

of Novel Corona virus. People (including all ages) can be affected by this
new type of virus. Till time there is neither vaccination has been invented nor
arrested by antibiotics due its unknown characteristics. Many researchers are
trying to develop medicines against it. At present self-protection is very much
essential to combat with Corona. That is why disinfection/sanitization is one
important aspect too. In most places liquid sanitizer like Sodium hypo chloride
are in use. But in indoor application liquid sanitizer can affect or deformed
the materials. From this point of view one can utilize photo biological effect
of light. UV range of the light source has established its ability to destroy
bacteria and virus too. Experimentation had been carried out in Illumination
Engineering Laboratory of Jadavpur University with several light sources with
384 Advances in Data Science and Computing Technology

different intensities of UV contents from light sources to reduce germicidal

effects. The process shows a successful result to reduce bacteria and viruses.
Same may be utilized for sterilization of indoor areas. But UV has some
deleterious effect on human being thus presence of them must be avoided. For
earlier studies on other viruses, like H1N1, Ebola, etc. UV with proper chemical
agents as a catalyst has found positive effect to combat viruses. The above two
factors have been discussed in details in this chapter.

26.1 INTRODUCTION

Light is a popular form of energy resource which primarily provides

visibility. If an energy resource can be harnessed properly same can be
utilized as a weapon for good or bad purpose. At present entire world is in
panicky situation due to sudden and wide spread attack of COVID-19 (Zaria,
2020), corona virus. Since it is an airborne virus, entire localized region
must be sanitized properly by breaking / weakening the RNA structure of
the virus. Chemicals are introducing through spray to combat the same. But
in some confined places it is not possible to use the same as it may create
damage of the surrounding material, making it wet, at the same time bring
permanent health hazard to living beings. This calls for a device which may
be used without contaminating the surrounding other living and nonliving
objects. In this work effort has been made to develop a system to prevent the
environment from the virus as well as a way of survival from the disease too.

26.2 PROPOSED METHODOLOGY

Light is an electromagnetic wave. It is accompanied by ultraviolet ray (UV)

(Bolton, 2010) and infrared ray (IR) in Electromagnetic spectrum. Every­
body gets scared by the word UV. But as said earlier proper knowledge
helps us to harness as well as divert the same in proper way. The above
waves have been further classified according to its characteristics and
analysis (Table 26.1).
• VUV (Vacuum UV) light is harmful due to its ability of immediate
retort organic molecules even at low doses and with oxygen atoms.
• UV-C light has biocide effects and usually utilized as germicidal agent.
UVC directly affects deoxyribonucleic acid (DNA) and ribonucleic
acid (RNA) by inducing molecular transformation (i.e., producing
photoproducts in the genetic material).
A Proposal for Disinfection of Contaminated Subject 385

• UV-B light: Photons in this range are known for sun burning, leads to
photo carcinogenesis and photo aging.
• UV-A light being comparatively longer in wave length has less detri­
mental effects.

TABLE 26.1 Range of Electromagnetic Spectrum (Bolton, 2010)

Wave Length (nm-Nanometer (10–9), Unless Stated Name of Wave
100–200 nm Vacuum UV (VUV)
200–280 nm UV C
280–315 nm UV B
315–380 nm UV A
380–780 nm Visible light
780 nm–1.4 µm IR A
1.4 µm–3.0 µm IR B
3.0 µm–1.0 µm IR C

During time of survey, it was observed that most of the solid materials
like paper documents, textiles materials, etc., becomes brittle in nature
and changes its color as it ages. It is well known fact that microorgan­
isms (Kowalski, 2009) may be responsible for the same. Same is also
observed in water and air medium which are affected by unwanted growth
of microorganism. So, at first identification of the types of micro-organism
are needed which are grown on subject materials. If micro-organism is
identified by experimentation, then it can be tries to remove or control for
safety purpose.
Some microorganisms were identified through microscope in laboratory
studies at Jadavpur University (Table 26.2).
As an illumination engineer it is necessary to find out ways and means
for destroying microbes like fungi, algae, bacteria, etc., with the applica­
tion of light. An effort has been made by using UV radiation which is
available in artificial light sources specially mercury vapor discharge type
light sources.
In this chapter various types of lamps, i.e., incandescent lamp, Conven­
tional Fluorescent Tube, CFL, and LED (both warm and cool white) and
UV-C lamps are experimented to find out the UV-C ranges in it (Table 26.3).
Experimentation has been carried out to measure the amount of UV-C
of the lamp which is mounted horizontally from top, facing downward. The
test results at various distances are as follows (Table 26.4 and Figure 26.1).
386 Advances in Data Science and Computing Technology

TABLE 26.2 Few Common Microorganism and UV Needed to Combat (Philips Catalogue,
n.d.)
Microorganism Under Discussion Dose Needed to Kill with UV (J/m2)

Mold Spores

Aspergillus niger 1,320

Rhizopus nigricans 1,110

Penicillium digitatum 440

Virus
MS-2 Coliphase 186

Rotavirus 81

Hepatitis A 73

Polio virus 58

Influenza virus 36

Bacteria
Salmonella typhimurium 80

Bacillus anthracis 45.2

Vibrio chlolerae 35

Staphylococcus aureus 26

TABLE 26.3 UV-C Output from Various Artificial Light Sources

Distance Lamp Specification with UV-C (Microwatt/cm2) Range
(ft.) Incandescent FTL CFL Cool CFL LED LED UV-C
(40 Watt) (40 watt) (8 watt) Warm Cool Warm Lamp
(8 watt) (8 watt) (8 watt) (8 watt)
2 2.1 0.9 0.6 0.5 0.3 0.3 91
4 0.9 0.2 0.3 0.3 0.2 0.1 38
6 0.7 0.1 0.2 0.1 0 0 18
8 0.5 0 0.1 0.1 0 0 7

Mathematical modeling through MATLAB simulation can be done as

follows:
 Along the Vertical Axis from Lamp:
UVver = 584.08e–0.454x, Where x is the distance of separation.
Thus the generalized equation can be drawn as UVver = Av × e–bvx.
 Along the Horizontal Axis from Lamp:
UVhor = 239.27–0.455x, Where x is the distance of separation.

Thus, the generalized equation can be drawn as UVhor = Ah × e–bhx.

A Proposal for Disinfection of Contaminated Subject 387

TABLE 26.4 Distance vs. UV-C Ranges of Test Lamp

Position SL. No. Distance (ft.) UV-C (µW/cm2)
Vertical 1 1 153
(from the center of the 2 2 95
lamp) 3 3 57
4 4 40
Horizontal (from the center 1 1 148
of the lamp) 2 2 91
3 3 55
4 4 38
5 5 27
6 6 18
7 7 11
8 8 7
9 9 4
10 10 2

FIGURE 26.1 Nature of curve (decaying) for UV-C with distance.

The nature of decay of UV-C is same and values of “ bv” and “ bh” attained is
almost same. Values of “Av” and “Ah” are not same as the through of light is not
the same in all directions due to its linear shape. The intensity of factor “Av” and
“Ah” will vary with the type, size, shape, and intensity of radiations of the lamps.
Thus UV-C lamp employed for this experimental purpose was T5 Slim
Line of 4 W, 6,400 K, 306 microwatt/m2 output.
This experiment attempts to give a brief idea about how UV radiation
destroys or control microbes like fungi, algae, and bacteria with specific kind,
intensity, and exposure time.
388 Advances in Data Science and Computing Technology

26.3 EXPERIMENTAL STEPS

The experimental steps are shown in Figure 26.2.

FIGURE 26.2 Procedure of collection of microorganisms and reduction by UV application.

26.4 NECESSARY INSTRUMENTS TO CARRY OUT EXPERIMENT

1. Laminar Flow Meter: This apparatus is laboratory-based equip­

ment used for germicidal operation incorporating UV lamps. This
A Proposal for Disinfection of Contaminated Subject 389

machine is not in use to keep the interior of the work zone clean and
decontaminated.
2. Autoclave: It is a steam sterilizer machine. It is mainly use steam
under pressure to kill harmful microorganism like bacteria, viruses,
fungi, and spores. The items are heated to an appropriate sterilization
temperature for a specific amount of time.
3. Incubators: Incubator is a device which is used to grow and maintain
cell or microbiological cultures. This instrument maintains humidity,
optimal temperature, and other constraints (Figure 26.3).

FIGURE 26.3 Incubator chamber.

26.5 EXPERIMENTAL DATA

From our earlier laboratory studies at Jadavpur University—under School of

Illumination Science and Engineering (SISED) in collaboration with School
of Environmental Studies of Jadavpur University upon Bacteria results are
as follows:

26.6 OBSERVATIONS

For the above studies it has been observed that as the intensity and duration
increase effectiveness of reduction for bacteria’s colony increases for both cases:
 390
1. UV Treatment Using Swab Test (Table 26.5):
TABLE 26.5 Result of Swab Test Method (Anubrata and Kamalika, 2019)
Sample Initial No. Intensity Duration Dose Final no. of Reduction Observations
of Bacterial of UV (hr.) (mJ/cm2) Bacterial (%)
Colonies lamp colonies
I.A 15 612 3 1.836 6 60.00*a Time remaining same with double intensity reduction
I.A 15 306 3 0.918 8 46.66*a is only 15%
I.A 60 612 6 3.672 17 71.66*b By increasing the Duration of time with higher

Advances in Data Science and Computing Technology

intensity further 11.66% improvement
I.A 60 306 6 1.836 31 48.33*c Not much effective with lower intensity
I.B 9 612 3 1.836 3 66.66 *d
Time remaining same with double intensity reduction
I.B 9 306 3 0.918 5 44.44*d is 20% in case of lower initial concentration
I.B 15 612 6 3.672 4 73.33*e By increasing the Duration of time with higher
intensity further 10% improvement
I.B 15 306 6 1.836 7 53.33*f Not much effective with lower intensity
II 17 612 3 1.836 6 64.70 *g
Time remaining same with double intensity
II 17 306 3 0.918 10 41.17*g reduction is more than 20% in case of lower initial
concentration
II 25 612 6 3.672 6 76.00*h By increasing the Duration of time with higher
intensity further 11.3% improvement
II 25 306 6 1.836 13 48.00*i Not much effective with lower intensity
 A Proposal for Disinfection of Contaminated Subject
2. UV Treatment Using Replica Platting Method (Table 26.6):
TABLE 26.6 Result of Replica Platting Method (Anubrata and Kamalika, 2019)
Sample Initial no. Intensity Duration Dose Final no. of Reduction Observations
of Bacterial of UV (hr.) (mJ/cm2) Bacterial (%)
colonies lamp Colonies
I.A 37 612 3 1.836 14 62.16*a Time remaining same with double intensity
I.A 37 306 3 0.918 19 48.64*a reduction is only 13.5%
I.A 60 612 6 3.672 15 75.00*b By increasing the Duration of time with higher
intensity further 11.66% improvement
I.A 60 306 6 1.836 28 53.33*c Not much effective with lower intensity
I.B 29 612 3 1.836 10 65.51 *d
Time remaining same with double intensity reduc­
I.B 29 306 3 0.918 15 48.27*d tion is 17% in case of lower initial concentration
I.B 55 612 6 3.672 12 78.18*e By increasing the Duration of time with higher
intensity further 13% improvement
I.B 55 306 6 1.836 25 54.54*f Not much effective with lower intensity
II 21 612 3 1.836 7 66.66 *g
Time remaining same with double intensity reduc­
II 21 306 3 0.918 11 47.61*g tion is 20% in case of lower initial concentration
II 33 612 6 3.672 7 78.78*h By increasing the Duration of time with higher
intensity further 12.12% improvement
II 33 306 6 1.836 16 51.51*i Not much effective with lower intensity

391
392 Advances in Data Science and Computing Technology

• Time remaining same with double intensity reduction is only 13 to 15%

• By increasing the Duration of time at higher intensity further 12%
improvement.
• Not much effective with lower intensity.
• Time remaining same with double intensity reduction is 16 to 20% in
case of lower initial concentration.
• By increasing the Duration of time with higher intensity further
10–12% improvement.
• Not much effective with lower intensity.
• Time remaining same with double intensity reduction is only 20%.
• By increasing the Duration of time at higher intensity further 11–12%
improvement.
• Not much effective with lower intensity.

26.7 RESULT ANALYSIS

Conventional mercury vapor discharge lamps are rich source of UV. Even
newly introduced. LED lamps also content UV-C. Thus, an array of UV
lamps can be able to sanitize indoor area. The arrangement can be designed
based on the type, size, and density of virus as well as size of the place, etc.
Since UV-C is carcinogenetic for the long time exposure and it can damage
eyes restriction of human movement in the place must be maintained.
From our earlier studies it appeared that during ancient time people
in Egypt used to treat through Phototherapy by introduction of UV-A and
bio extract – Psolaran (Ammi majus). At present synthetically developed
Amotosalen (S-59) along with UV may alleviate respiratory suffocation.
Here TiO2 has been reported to act as a catalyst.
From earlier studies it has also been found that UV has been successful
upon H1N1, Ebola, etc., virus with UV irradiation of 2 mJ/cm2 of 222 nm and
4 J/m2 (0 to 30 s) of 254 nm, respectively.

26.8 CASE STUDIES (DAVID, N.D.)

Under the guidance of David Brenner, PhD, Director of the Center for
Radiological (Columbia University) believes they may have identified a
new, low-cost solution to eradicating airborne viruses in indoor public areas.
Here reports has been attached in Table 26.7.
 A Proposal for Disinfection of Contaminated Subject
TABLE 26.7 Some Information from the International Study (David, n.d.; Misc, n.d.)
Disease Causative Agent Organism Symptoms Effective Infective Dose Incubation Mortality
Persistence on Person (Aerosol) Period
to Person
Anthrax Bacillus anthracis, Life of Spores Fever, cough, No
10,000– 50 000 1–6 d High if
infected animals may be more discomfort in
spores symptoms
are responsible than 40 years respiratory track,
occur.
for transmitting to etc.

humans.
Plague (i) Bubonic, Yersinia pestis Can persist in Cough and cold Yes
<100 organisms 1–10 bacteria, High, unless
(ii) Pneumonic, soil for about followed by High
more for treated
24 days. fever, Dyspepsia, etc.
aerosol immediately.
(iii) Septicemic (100–20,000)
Treatment
Ebola virus disease Filoviridae family, Stable Intense weakness, Yes
i) particles 1–21 d, 90% fatality
(i) Bundibugyo virus, genus Ebolavirus muscle pain,
of < 5–10 μm avg. 8–10d
(ii) Ebola virus, caused by plants, headache, sore throat,
ii) droplets
(iii) Sudan virus, arthropods, birds, vomiting, diarrhea,
of diameters
(iv) Taï Forest virus etc. rash, impaired kidney,
> 20 μm
and liver functions

393
394 Advances in Data Science and Computing Technology

26.9 CONCLUSION

From the above studies it can be inferred that UV can be utilized in controlled
manner to sanitize the infected areas and large scale water purification
(Adrita et al., 2020) for rural area. Further studies can also be made with
Photo therapy along with chemical aids as stated above for the diseases
where Antibiotic fails. In this case development of catalyst with the help of
upcoming Nanotechnology will be a great solution.

KEYWORDS

• COVID-19 virus
• germicidal effect
• photo biological effect
• ultraviolet

REFERENCES

Adrita, S., Antalina, S., Soumya, G., Ratan, M., & Kamalika, G., (2020). “Development
of UV aided water purification system powered by Solar energy for remote villages.”
International Conference on Sustainable Water Resources Management Under Changed
Climate. Organized by School of Water Resources Engineering, Jadavpur University,
Kolkata in Collaboration with Department of Geography, Women College Calcutta and
NIT Durgapur.
Anubrata, M., & Kamalika, G., (2019). “Studies on germicidal benefit of ultra violet ray upon
old paper documents.” Light& Engineering Journal, 24(4), 17–24. ([ISSN 0236-2945].
Scopus Indexed & SCI, Peer Reviewed & UGC Approved. (Included UGC Care List:
Group-A).
Bolton, J. R., (2010). Ultraviolet Applications Handbook (3rd edn.). (Updated) // Bolton Photo
sciences Inc.
David, B. Misc. publications of PhD, Director of the Center for Radiological Research and
Professor at Center for Radiological Research at Columbia University Irving Medical
Center. http://www.columbia.edu/~djb3/ (accessed on 11 January 2022).
Kowalski, W., (2009). Ultraviolet Germicidal Irradiation Handbook UVGI for Air and
Surface Disinfection (p. 501). Springer Heidelberg Dordrecht London New York.
Misc. publications of National Library of Medicine, US. https://collections.nlm.nih.gov/
(accessed on 11 January 2022).
A Proposal for Disinfection of Contaminated Subject 395

Philips Catalogue, Philips lighting B.V. UV Health and Wellness. www.philips.com/uvpurifi­

cation (accessed on 08 December 2021).
Zaria, G., (2020). BBC survey. “There’s Only One Type that Can reliably Inactive COVID19­
and it’s Extremely Dangerous.” https://www.bbc.com/future/article/20200327-can-you-kill­
coronavirus-with-uv-light (accessed on 11 January 2022).
INDEX

A applications, 220

automation, 153

Abnormal

backgrounds, 187

conditions, 198

sector, 188, 220

overheating, 271

Agro nutrition, 188, 194

Absolute mean brightness error (AMBE), 3,

alert (ANA), 1, 69, 71, 74, 78, 79,

8, 9, 12, 13, 16

187–190, 194

Accelerometer chip, 148

Accounting
Ailment symptoms, 16

information, 350–353, 357, 360

Alert

process, 349, 350, 359, 360

center, 158, 159

professionals, 349, 352, 359

message, 158, 159, 161–164

software, 349–352, 355–360

Algorithm, 5, 21, 24, 28, 30, 31, 36–38, 40,

Accuracy, 9, 14, 15, 26, 28, 29, 60, 62–64

41, 49, 50, 59, 60, 62, 71, 74, 100, 102,

handwritten text recognition tool, 20

107, 109, 111–113, 116, 119–124, 126,

phase-shift information, 205

132–135, 139, 200, 201, 227, 228, 254,

value, 19, 29, 32

277–281, 284–287, 294, 300, 305–310,

AC-DC-AC conversion, 260, 262

319, 324–330, 335, 336

Achievability, 188
Alkaline chemistry, 157

Activation function, 11, 27, 64, 74, 76

Amazon, 87, 88, 92, 93, 96, 110

Actual
Ambient temperature, 238

gradient value, 24
Ambulance, 158, 159

non-vessel pixels, 9
Amusement park, 96

raw pixels, 60
Analysis

vessel pixels, 9
brushless DC motor, 229, 234

Adagrad, 24, 28–32

DC-DC converter, 229, 231

Adam optimizer, 24, 28, 74

discussions, 89, 103

Adaptive
expanded examination revelation

cruise control (ACC), 291

(noteworthy bits of knowledge), 89

gamma correction (AGC), 4–7, 13–17

improved client encounters purchaser

histogram equalization (AHE), 4, 6

ventures, 90

Additive white gaussian noise (AWGN),

simpler progressively secure

335
approaches (pay on the web), 91

Adenine, 322
solar pv array, 228, 230

Adequate irrigation, 188

Analytical

Ad-hoc
calculations, 291

networks, 290–292, 294

E-commerce, 93

vehicle network, 289

Anatomical artifacts, 4

Administrative bodies, 48
Anchor

Advanced technologies, 146, 165

node, 69–73, 75–78

Affordability, 185
street, 291

Agricultural
Android

398 Index

application, 148
Authentication, 276, 283, 285

interface, 148
problem, 20

Anti
Authorized

aliasing technique, 25
user, 276, 306, 307, 316

allergy drops, 170

vehicles, 290

antibiotic drugs, 170

Auto-encoders, 60

jamming authorization, 334

Autoclave, 389

Application oriented, 5, 19
Automated, 188

network architecture, 22
disease analysis system, 3

Arbitrary
learning algorithms, 21

combinations, 17
perception, 59

design, 364
retinal

Architecture, 63
analysis system development, 4

Arduino, 152, 153, 159, 164–166, 175, 177,

diagnostic system, 4

178, 180, 181, 184, 190, 193, 194

image analysis system, 4

boards, 152, 180

torque, 257

IDE, 152, 178, 190

Average

micro-controller, 166
accuracy, 69, 73, 74, 76, 78, 79

board, 152
range error, 69, 71, 73, 74, 77, 79

nano, 164, 178, 180

training, 14

ultrasonic sensor techniques, 175

accuracy, 14

uno microcontroller, 165

Arithmetic
B
coding, 330
Background diabetic retinopathy, 10

encoding
Bacteria, 383–385, 387, 389

algorithm, 327
Ball bearings, 377

technique, 320, 327

Barker code, 334, 336

Arthritis, 172–175, 183

Base

Artifacts, 4, 62
isolation, 377

Artificial
technique, 378, 379

feed-forward neural network, 70

station (BS), 70

intelligence, 36, 44
Baseline intraocular pressure, 173

light sources, 385

Batch size, 12, 22, 23, 25, 29–32, 63

neural network, 22, 79

Battery backup, 158

Assessment procedure, 183

Bayes net, 21

Asymmetric
Bearing fault, 204

algorithm, 285
Behavioristic, 85

cryptography, 327
Better predictive analytics, 86

key, 278, 320, 323, 325, 330

Bi-directional communication, 102

algorithm, 278, 323

Big data, 81, 84, 85, 89, 90, 92, 93, 95–105,

cryptographic scheme, 325

320

cryptography, 309, 323

analytics, 84, 85, 89, 93, 103

encryption, 320
amalgamation, 89

Athletes, 148
companies, 84

Atrophy, 10
environment, 97

Attractive dashboards, 102

solutions, 103

Audible range, 180

validation tools, 97

Audio steganography, 308

Bilinear interpolation, 6

Index 399

Bio-medical images, 5
Cell state, 52

Biocide effects, 384

Center

Biological
average methods, 247

DNA sequences, 320

gravity (COG), 247

molecules, 320
Centroid defuzzification, 247

Biomedical image data enhancement, 7, 73

Ceramic antenna, 152

Bit error rate (BER), 334, 337–339, 344

Change error (CE), 61, 246, 247

BLDC motor, 219–222, 227–229, 234, 235

Channels of communication, 277

Blindness, 171
Character

Blinking LEDs, 38
classification methods, 20

Blood vessels, 4
recognition, 21, 22

Body temperature (BT), 145, 148, 149

Charging-discharging, 226

Boltzmann constant, 241

Chemical

Bond degradation, 379

analysis, 189

Boost converter, 239, 243

fertilizers, 187, 188

Bottom-hat transformation, 8
laboratory, 187, 189, 194

Braced frames, 378

Child trafficking, 162, 164

Broadcasting, 302
Chinese remainder theorem, 320, 321, 325,

Broken rotor bar, 204

327, 328, 330

Business
Chip payment cards, 276

analytics, 97
Chronic problem, 170

atmosphere, 351
Cipher text, 277, 281, 283–285, 305, 306,

brand value, 84
308, 316, 324, 328

communications, 276
Class investigation center point, 86

development, 87
Classification

endeavors, 91
accuracy, 19

environment, 96, 97
efficiency measurement, 10

fundamental concern, 91
evaluation, 5

opportunities, 97
quality assessment (CQA), 8

performance, 97, 351
Classy fuzzy logic MPPT controller, 237

plan, 165
Client, 83–87, 92, 93, 109, 110, 119, 120,

transactions, 95
131, 278, 283, 358

Buttery maculopathy, 10
dedication, 84

Buyer demand pattern, 120

execution variable, 84

Buzzer, 179–181, 184

hello, 277, 278

Bygone models, 60
Climatic parameters, 188

Cloud, 86, 109–112, 119–123, 127, 131,

C 132, 134, 140, 158, 163, 166

Cake pops, 86
architecture, 134

Canny edge detector, 41, 43

computing, 109–112, 119–123, 127,

Capacitor, 223, 226, 244

131–134, 140

equivalent series resistor, 226

background, 111

Capillary action technique, 174

classification of load balancing tech­
Carbon-zinc, 157
nique, 121

Carcinogenetic, 392
cloud system, 121, 132

Cardio-vascular diseases, 4
load balancing, 121, 132

Carrier media, 316

data centers, 120

Causation multiple messages, 290

environment, 112

400 Index

system, 110, 132

Computational

technology, 121
characteristics, 321

Clustering, 4, 37
complexity, 22, 29, 31

Coaching classes, 162

science, 276

Coal industry, 70
software, 204

Coast guards, 159, 163

time, 65

Code division multiple access (CDMA),

vision, 35–38, 40, 41, 44, 59, 60, 63, 149

333, 334, 338–344

algorithms, 35

Coefficients of,
methods, 41

correlation, 199
techniques, 37

wavelet, 200
Computerization, 188

Coif, 199, 200, 204

accounting, 357

Coiflets, 199, 200, 204

Conditional random fields, 5

Color histogram, 60
Confidential, 276, 277, 328, 330

Colossal data, 89, 90

data, 316

Column
Confusion matrix, 62, 64

beam joints, 378

Congestion filter, 257

sum, 112, 113

Congruent equation variable, 321

Combat viruses, 384

Connectionist temporal classification

Commercial companies, 176

(CTC), 20

Communication
Consecutive nodes, 23

channel, 275, 278

Constant voltage methods, 227

experience, 98

Consumer

network, 306, 307

loyalty, 84

process, 284

oriented applications, 132

science, 276

Contemporary literatures, 3

Communities, 100

Content analysis, 97

Community transmission, 48, 49

Continuous conduction mode, 229, 231

Compact gadgets, 84

Comparative analysis, 21, 62, 97, 119

Contrast

Competent
enhancement, 6, 8, 73

network, 19
limited adaptive histogram equalization

scheduling technique, 127

(CLAHE), 4–6, 12–17

Competitive market, 95, 97, 103

Control

Completion times, 126, 135

input values, 246

Complex
pitch accurator system (CSA), 257

variants, 60
Convenience of research, 353

vision problems, 60
Conventional

Complimentary
DC motors, 220

delivery, 86
energy

transportation, 87
crisis, 237

Comprehensive
resources, 237

eye examination, 173

fluorescent tube (CFL), 385, 386

wireless network, 289

mercury vapor discharge lamps, 392

Compression
motors, 220

algorithms, 277
neural network, 61

permutation, 280
Converter, 219, 221–229, 231–235, 238,

spring, 182
239, 243, 244, 249, 250

technique, 327
duty cycle, 239

Index 401

Convolution
Cyber-crime, 306

layers, 11, 61
Cytosine, 322

neural network (CNN), 4, 5, 10, 11, 13,

15–17, 21, 60–62, 66, 67

D
architecture, 10, 13

Dashboard camera, 36

Cornea, 175

Data

Correlation

accessibility, 99

analysis, 202, 204

acquisition

coefficient, 198, 199, 202–212, 215

system, 204, 205

technique, 202

task, 205

Corresponding output matrix, 76

analytics, 85, 102, 105

Cost reduction, 103

augmentation, 12, 63

Country-wide prediction, 49

bank, 84

Covariance matrix, 202

breaches, 105

Cover writing, 307

centers, 120

COVID-19, 47–50, 52, 53, 55, 56, 255, 384,

encryption, 275, 287, 316, 330

394

encryption standard (DES), 275, 280,

virus, 48, 52, 394

281, 284–287, 305, 306, 316

Creo elements, 182

algorithm, 284, 286, 305, 306

Critical bending moment, 380

decryption algorithm, 281

Cross-cutting investigation applications, 85

innovation, 92

Cross-entropy, 61

management, 99

loss function, 11

preprocessing, 52

Cryptanalysis, 308

reception, 70, 72

Cryptographic, 276–278, 292, 305, 308,

sets, 50, 60

309, 319–325, 327–329

traffic, 96, 103, 285

algorithms, 282

transmission, 70, 72, 277, 298, 320

asymmetric key cryptography, 309

Database, 5, 21, 25, 26, 149, 158, 161, 162

hashing, 309

keys, 277, 278, 282

workstation, 111

symmetric key cryptography, 308

Dataset, 9, 10, 12, 16, 17, 21, 25, 36, 50, 52,

Cryptosystem, 308, 320, 321, 323, 326

62, 64–66, 75

using knapsack, 330

description, 9, 25

Cumbersome, 105, 349, 350

DC

Cumulative distribution function (CDF), 6

DC

Current
boost converter, 238, 239, 243, 244,

input, 51, 52
246, 248–250

ripple, 225
converters, 220, 239, 243

signals, 197, 199, 202, 239

link capacitance, 226

signature analysis, 197

motor, 181

Customary datasets, 85
type electrical energy, 238

Customer, 85, 89–91, 96, 98–103, 123, 132,

Dead load (DL), 366, 369

350–352, 356, 358, 360

Debugging, 153

feedback, 102
Decentralized, 302

requirements, 99
Decision making, 38, 93, 349, 351, 360

satisfaction, 98, 350–352, 358

process, 95, 97

tendencies, 90
Decryption, 275, 277, 279, 281, 282, 284,

value proposition, 98
305–308, 310, 316, 320, 322, 323, 325,

wellbeing, 98
327–329

Customized recommendations, 87
process, 323, 325, 328

402 Index

Deep
key, 284, 286

architecture, 5, 19, 22
exchange algorithm, 275, 281, 286

classification
exchange protocol, 278

evaluation metrics, 5
Merkle algorithm, 278

experiments, 5
Digit

metric evaluation, 13
recognition, 19–22, 31, 32

process, 19
system, 21

ConvNet, 62
Digital

learning, 5, 12, 17, 20–22, 27, 29, 35–40,

currencies, 276

44, 59–62, 66
image processing, 6

application, 20
India, 56

architecture, 5, 22

object recognition, 59

environment, 20

people, 101

methods, 21

retinal image for vessel extraction, 9

model, 39, 40, 60, 66, 67

result verification, 17
technology, 99

techniques, 21
Digitization, 20, 99

neural network (DNN), 5, 10–12, 19, 20,

Digitized

22, 31, 32, 61

handwritten documents, 20

architecture, 11, 19
system, 99

implementation requirements, 12
Dilating drops, 170

training, 12
Diode (D), 244, 246, 247, 253, 258, 285,

structural benefits, 61
286, 289, 292, 299, 323, 363, 370, 371,

Defense sector, 165

392

Defuzzification, 245, 247

Direct sequence spread spectrum (DSSS),

Degree of,
333, 334, 344

accuracy, 49
CDMA

acidity, 189
communication system, 334

agreeability, 353, 355

receiver system, 334

Delivery, 98, 99, 101, 109, 131, 173, 175,

Directional greedy routing (DGR), 291

219, 291
Discrete

channel, 101
logarithmic

Demand forecasting, 101

calculations, 286

Demographic detail, 101

problem, 284

Denoising, 203
wavelet transform (DWT), 197, 199–201,

performances, 199

215

Deoxyribonucleic acid (DNA), 319–323,

Disintegration, 36

327–330, 384

Disk shaped masks, 7, 73

cryptography, 320, 321, 327, 328

Distant objects, 149

encoding, 320, 327, 330

Dependability, 92, 123

Distinctive assortment, 83

Detection, 21, 37, 39, 59, 60, 62, 103, 178,

Distorted, 61, 339, 344

189, 197, 198, 215, 301, 334, 335

continuous signal analysis, 199

Detrimental effects, 385

stator currents, 199

Device detection range, 178

Distraction of mind, 184

Diabetes, 3
Distributed

retinopathy, 10
channel activities, 99

Diffie-Hellman
nature, 290

algorithm, 278, 279

system, 120, 123, 127, 133

design, 284, 285

District laboratory, 190

Index 403

Documents study, 97
Efficiency, 4, 5, 16, 19, 20, 24, 36, 38, 41,

Domain, 5, 19, 20, 32, 70, 103, 197, 200,

61, 71, 103, 110, 164, 219–222, 234, 235,

202, 203, 250, 320, 334

267, 271, 284, 358–360

DoS attack, 290

Elastic modulus, 364

Drinking, 39
Electric

Drive, 9, 10, 12–16, 219

circuits, 152

train model, 258

consumption, 178

Driver attention
devices, 271

monitor, 41
fault, 199

tracking, 38
grid network, 256

Drop bottle holder device, 169, 172

power, 254

Dropbox, 110
science, 276

Dropout, 22, 23, 25, 27, 29–32

torque, 260

hyperparameter, 25
vehicles (EVs), 239, 250

value, 23, 27, 29, 31, 32

charging stations (EVCSs), 239

Drug, 170–172, 174–177, 182, 184, 185

Electromagnetic

installation, 172, 176

radiations, 189

schedule, 172
spectrum, 384

Dry eye symptoms, 170

torque, 229, 234

Ductility, 378, 379

wave, 384

Durability, 170, 183

Electron

Duration of time, 392

charge, 241

Duty cycle, 239, 243, 244, 249, 250

design, 184

Dynamic
Electronic

algorithm, 111, 121, 122, 133

aids, 184

load balancing algorithms, 111

arduino, 185

performances, 234
commerce, 84, 276

state performance, 229

commutation, 220, 221, 227, 228, 231

method, 228

E components, 155

devices, 84, 183, 220

Earthquakeresisting structures, 364

gadgets, 84

Ease of implementation, 234

model, 155, 170, 175, 182

Easy eye dropper, 172, 175–177, 182, 185

technology, 148, 170

E-commerce, 93
Elliptic curve, 285

business, 83
discrete logarithm problem, 286

organizations, 85
E-mail, 148

sites, 83
Emergency

websites, 89
alert systems, 165

Economic
responders, 149

condition, 101
Empirical

indicators, 101
research, 97

operation, 253
studies, 85

Edge
Encoded encrypted image, 309

detection, 37, 60
Encrypted, 275, 277–282, 284–286,

node based greedy routing (EBGR), 290

305–308, 310, 316, 319, 320, 322, 323,

segmentation, 37
325, 327, 328

Effective enhancement technique quantita­ data, 307, 320, 327

tively, 5
message, 310

404 Index

purpose, 323
dropper, 169, 170, 173, 176, 178, 180

End-to-end latency, 290

examinations, 170

Energy
lids, 175

consumption, 122
piece arrangements, 176

dissipating, 378

devices, 378, 379

F
seepage, 255
Facial recognition, 60

storage unit (ESU), 239, 243, 249, 250

False

Enhancement techniques, 3, 5, 8, 12, 15–17

negative (FN), 9, 26, 63

Enterprises, 103

positive (FP), 9, 26, 63

Entity coordination, 292

recognized non-vessel pixels, 9

Environmental parameters, 70

Fast fourier transform (FFT), 197–200

Epicenter, 364

Fault

Epochs, 14, 21, 32, 63–65, 67

detection, 197, 200

Error

diagnosis, 199

backpropagation algorithm, 61

prediction, 198

matrix, 76

tolerance, 110, 122

percentage, 56

Feasibility study, 189

prone, 20

Evaluation metric, 5, 8, 9, 14, 19, 26

accuracy, 190

deep classification metrics, 9

cost, 189

IQA metrics, 8
lifetime, 190

Eve teasers, 159

time consumption, 190

Execution measurements, 83
Feature extraction, 203

Existence of scaling function, 203

Fiber

Expansion
confinement, 379

boards, 152
reinforced concrete, 379

permutation, 280
Final

Experimental
cipher text, 327

phase, 59
permutation, 281

set up, 204

Finance, 96

data acquisition system, 205

information, 352, 359, 360

Explicit node, 123

portfolios, 132

Explorative study, 86
statements, 349, 350

imagining future Wal-Mart, 88

Finest scale approximation, 200

Wal-Mart versus the opposition, 87

Fine-tuned layers tested, 61

Exponential
Fire

dampers, 378
breakout, 163

growth, 104
brigade, 158, 163

External
fighters, 160, 161, 163

attacks, 292
Firmness abnormality, 367

communication, 356, 357

Firmware, 153

Extraction process, 200

Fitting explicit requirements, 83

Eye
Flora micro-controller, 165

ball center, 173

Forefront advancements, 91

drop
Foreseeing conduct, 83

applicator, 173, 174

Forget gate, 50, 51

bottle holder, 170, 172, 175

Forums, 100

installation, 171
Fossils energy, 253

Index 405

Fovea, 4, 10
system for mobile communications

Frequency
(GSM), 148, 149, 153, 154, 164, 166,

bandwidth, 198
190

domain signal, 203

module, 149, 154, 164

information, 198
warming, 238

Friction bearings, 377

Wind Energy Council (GWEC), 255

Fungi, 385, 387, 389

Google

Fuzzification, 245, 247

drive, 110

Fuzzy
online colab cloud environment, 12, 27

control, 245
tracks customized ads, 87

logic controller (FLC), 238, 245, 250

trends online tools, 84

MPPT, 245
Government agencies, 44, 105

controller, 238, 239, 245, 246, 248, 249

Gradient

de-fuzzification, 247
descent, 24

fuzzification, 245
exploding problem, 50

inference engine, 246

Gray scale, 40

image, 40

G Greedy traffic-aware routing, 291

Green channel images, 12

Gaming, 112
Greenhouse

Gamma
farming, 188

correction, 4, 6, 12
gas emission, 238

Gate signals, 228

Grid

Gaussian-filtering, 37
frequency, 254

Generators, 254, 255, 257, 262, 285

network, 257, 271

Genetic material, 384

operation, 255

Geographical
reliability, 254

routing protocol, 291

Ground movements, 364, 367

statistics, 363
Guanine, 322

Germicidal

agent, 384
H
effect, 384, 394
Hadoop, 97

operation, 388
Hall

Giant companies, 96
effect signal position sensors, 222

Gigantic data, 89
sensor signal values, 228

Glaucoma, 3, 171, 173, 182, 184, 185

Hand

sufferers, 171
tremors, 172

Global
written character, 20

big data market growth, 105

Handheld

business scenario, 358

devices, 84

casualty, 48
gadgets, 84

energy crisis, 220

Handpicked features, 22

environment, 254, 320

Handwritten

market, 88, 96, 254

character, 21

pandemic, 47, 48
digits, 25, 26

positioning system (GPS), 147–149, 152,

recognition, 20, 21, 32

164, 166
text recognition tool, 20

module, 152, 164

Harassment, 151, 162, 163

406 Index

Hardware requirement, 151

Hough lane detection, 40

9V battery source, 157

Household applications, 220

arduino UNO microcontroller, 152

Human

breadboard, 155
error, 35, 36

jumper wires, 155

experiences, 59

NEO 6m GPS module, 152

intervention, 165

SIM800L GPRS-GSM module antennae,

observer perception, 8

153
to-computer interaction, 147

switches, 156
visual perception based metric, 16

Harmful cryptanalyst, 319

Humidity, 389

Harmonic, 257
Hungarian method, 112, 113, 115, 116, 136,

changes, 257
139

Hashing, 309
Hydraulic shock absorbers, 378

Hassle-free supply chain management, 103

Hyper

Heads up display (HUD), 35, 37

parameter optimization, 19, 22, 27, 32

Health issues, 238

architecture, 19

Healthcare, 105
batch size evaluation, 29

Heart rate (HR), 148

dropout evaluation, 29

variability (HRV), 148

graphical representation (hyper param­
Heat, 70, 164
eter evaluation), 30

Hemorrhages, 4
learning rate evaluation, 27

Heritage computing, 351

number of hidden layers, 28

Heterogeneity, 134
optimization algorithm evaluation, 27

Heterogeneous
parameter, 20–23, 31

nodes, 79
tuning, 23

sensor systems, 37
serious economy, 92

Heuristic method, 245

tension, 3

Hidden

layer, 21–24, 27–32, 69, 70, 74, 76

I

state, 51, 52
ICMP messages, 293, 296, 298

High
Ideal customer profile (ICP), 84

chance of elliptic curve discrete (ECDH),

Image

286
acquisition, 306

end architecture, 20
classification, 39, 44, 59–61

frequency signal components, 201

data, 4, 5, 10, 12–14, 16, 17, 21

gain converters, 220, 221

dataset, 4, 5, 10, 12–14, 17

quality analysis, 60
enhancement, 3–5, 8–10, 12, 16

voltage applications, 243

enhancement techniques, 6

Higher image representations, 60

adaptive gamma correction (AGC)

Histogram oriented gradients (HOG), 38

enhancement, 6

Historical
contrast limited adaptive histogram

data, 100
enhancement, 6

information, 50
morphological operation enhancement,

Homogeneous virtual machines, 120

7

Hop information, 71
net data set, 39, 59

Horizontal displacement, 378

processing, 7, 9, 25, 35, 37, 40, 44, 62,

Horrendous incidents, 163

67, 73, 305, 306

Horticulture, 188
systems, 25, 306

Index 407

techniques, 40
Innovative eye drop, 172

quality
Input

assessment (IQA), 3, 5, 8, 11–13, gate, 50–52

15–17 image, 309

selection, 309
layer, 23, 24, 27, 69, 74

recognition, 60
Installation procedure, 174, 176

segmentation methods, 37
Instillation attempts, 173

steganography, 305, 306, 308, 316

Instant customer communication, 102

technique, 305
Instantaneous power, 228, 246

transformation, 4, 6
Institutional review board (IRB), 183

Immediate rescue operations, 162

Instructive assortments, 91

Implementation requirements, 5, 27
Insulation breakdown, 271

Improvisation, 185
Integrated infrastructure, 36

Incandescent lamp, 385

Integrity, 276, 328

Incentives, 98
Intelligence

Incremental conductance, 227

business purposes, 103

Incubators, 389
gathering, 277

Independent-installation, 185
transport systems, 291

In-depth understanding, 52
video processing, 60

Indian market, 174, 175, 184

Intensity, 6, 7, 12, 25, 178, 363, 364, 366,

Individual
387, 389, 392

class probability, 23
Interceptors, 285

parameter, 22
Intermediate computations, 23

Induction
Internal

generator, 257
attacks, 292

machine, 215
communication, 359

stator, 198
threaded passageway, 173

motor, 197–199, 202–206, 215, 220

Internet, 84, 86, 109, 131, 146, 147, 153,

Inductor, 223, 225, 229, 231, 244

305–307, 316, 319, 320, 351

Industrialization, 220
business firms, 83

Industries, 21, 44, 103, 105, 198, 220

thinking (IoT), 147, 319, 320, 330

Inference engine, 245, 246

Intra business communication, 350–352

Influenzas, 50
Inventory, 95, 99, 103

Information
Inverse discrete wavelet transform (IDWT),

biological system, 86
201

dashboards, 102
Inverter, 204, 220, 222, 262

development, 83
Investment decisions, 350

driven
Iron-oxide, 189

advertising, 92
Irradiation, 219, 228–231, 234, 235, 238,

markets, 92
239, 241, 242, 245, 248, 249, 392

level, 229–231, 234, 235

transferring module, 291

Isolated workstations, 110

Infrared (IR), 177–181, 184, 384, 385

Itchiness-relieving, 170

obstacle sensor module, 178

I-V characteristics, 241

photodiode, 178

radiations, 178

receivers, 178

J
transmitter, 178
Jamming condition, 333, 334

Initial permutation, 280

Jumpers connecting wires, 190

408 Index

K model implementation, 63

proposed work, 62

Kelvin, 241

training fine-tuning, 63

Kernel functions, 61

Lithium-ion, 157

Key exchange, 276–278, 280, 286, 308, 323

Live load (LL), 366

algorithm, 278

Load

K-means clustering, 37

balance, 107, 109, 111, 112, 116, 117,

Knapsack

119–123, 126, 127, 131, 133–135, 139,

algorithm, 325, 327, 328

140

private key, 328

min-min (LBMM), 112, 116, 123, 126,

public key, 327

127, 134, 135, 139

K-nearest neighbor (KNN), 21

stratagem, 123

balancers, 123

L resistor (RL), 244

Laboratory equipment, 388

torque, 229, 234

Laminar flow meter, 388

voltage requirement, 243

Large
Localization, 61, 70, 71

scale water purification, 394

Location

span structures, 364

detection, 178

Latency, 37, 40
tracker, 145, 146, 158, 162–165

Lateral drift, 370

device, 145, 146

Launching new products, 86

Logarithmic loss, 64

Leakage inductance, 221

Logistics, 101, 105

Leaning rate, 23
Log-loss curves, 30

Learning
Long short

algorithm, 22–24, 31, 59

term memory (LSTM), 20, 21, 50

rate, 22, 24, 27–32, 61, 63

time memory, 56

evaluation, 28
Long stretch accomplishment, 91

value, 24
Loss function, 24

Least significant bit (LSB), 308–310, 316

Low frequency signal components, 201

algorithm, 309
Low narrowband jamming, 334

Legitimate prerequisites, 92
Low voltage ride through (LVRT), 254, 262

Light, 8, 37, 70, 73, 134, 149, 163, 174, 178,

Lubricating drops, 170

185, 188, 192, 220, 254, 360, 383–385,

Lucidity, 352

387
Luminance pixels, 4

detection and ranging (LiDAR), 37

Luminosity, 188

Lightweight, 172, 183

Line segment extraction, 37

M
Linearity, 204
Machine

Linguistic
fault analysis, 199

diversity, 349, 350, 352, 359, 360

learning, 21, 22, 24, 25, 36, 37, 44, 56,

inputs, 245
61, 100–102

mélange, 350
algorithms, 102

variables, 246, 247

problems, 21

Liquidity analysis, 350

systems, 25

Literature review, 61, 96, 291

tool, 105

data preprocessing augmentation, 63

Magic wristbands, 96

dataset description, 62
Magnitude variations, 202

Index 409

Makespan, 134, 135, 139

Media sharing, 112

Making strategic decisions, 102

Medial canthi area, 175

Malformed font style, 21

Medical, 4, 9, 10, 44, 48, 53

Malicious node, 289, 293, 296, 299, 302

image segmentation, 10

isolation, 299
professionals, 4

Mammogram images, 5
Message

Mammoth, 49
authentication, 278

Man-in
communication, 285

middle attack, 278

Method, 113, 131, 136, 200, 305

the-middle attack, 285

current signal reconstruction (decomposi­
Manual intercessions, 188
tion of wavelet), 201

Manufacturing performance, 89
evaluation, 9

Market
selection of optimal mother wavelet, 203

basket analysis, 101

statistical analysis, 202

identification, 101
wavelet analysis, 200

penetration, 102
Methodology, 50, 157, 175, 293, 352, 384

segmentation, 101, 102

long short term memory (LSTM), 50

Marketing
Metric

decisions, 85
of authentication, 292

departments, 101
values, 5, 15, 16, 29

MasterCard, 87
Metropolitan city, 149, 150

Mathematical
Micro aneurysms, 4

algorithms, 277
Microbes, 385, 387

design, 220
Microbiological cultures, 389

manipulation, 201
Microcontroller, 146, 152, 153, 190

model, 50, 102, 250

Microgrids, 254

tools, 200
Microprocessor, 146

MATLAB, 195, 199, 206, 220, 221, 228,

Microsoft office, 110

234, 235, 241, 254, 257, 333–335, 337,

Migration time, 122

344, 386
Military communications, 276

Matplotlib, 52
Min time, 135

Maximum
Mines environment, 70

average sensitivity, 16
Mini-batch gradient descent, 24

permissible drift, 371, 373, 374, 376

Minimum

power point, 238

completion time (MCT), 112, 117, 123,

tracking (MPPT), 221, 227, 231,

126, 140

237–239, 243, 245–250

permissible values, 257

row-sum, 136–138
Mining, 44, 70

Mean
Min-min (MM), 112, 116, 123, 126, 127,

integrated systems, 99
134, 135, 139

of maximum (MOM), 247

algorithm, 134

squared error (MSE), 8, 203

schedule, 134

Measurement syringe, 183

algorithm (MM), 112, 123

Mechanical
technique, 112, 123

fault, 197, 198

Mirror arrangement, 183

detection, 215
Misaligned rotor, 204

stresses, 257, 261

Mobile, 36, 84, 157–59, 162, 164, 193, 290,

torque, 227, 258, 260

307

type faults, 215

ad-hoc network (MANET), 289

410 Index

Model
path routing technique, 293

accuracy rate, 62
Multiprocessing techniques, 38

design, 63
Multistoried

learning phase, 24
building, 160

Moderate damage risk zone, 366

structure, 378

Modern

cryptology, 276
N
retail business, 103
Naïve-Bayes, 21

Modified National Institute of Standards

Nanotechnology, 394

Technology (MNIST), 21, 25, 26

Narrowband jamming, 334, 337

database, 25
environment, 334

Modular transformations, 327

Natural damping, 379

Moisture sensor, 190

NavShoe, 149

Molecular transformation, 384

device, 149

Momentresisting frame, 378, 379

Negative

Momentum, 24
big (NB), 246, 247

Monolithic prediction, 49
small (NS), 246, 247

Morphological
Neighborhood

enhanced, 5, 16
climate deviations, 86

abnormal images, 15
nodes, 293

image, 16
Network

techniques, 5
architecture, 20, 27, 31, 63

test images, 16
communication, 299

operation, 3, 4, 7, 16, 73
execution resources, 12, 27

operators, 4, 12
failures, 164

Morphology, 13, 17
hyper parameter, 23

Mother wavelet, 197, 198–200, 202–204,

optimization, 31, 32

206–212, 215
learning, 22

selection, 215
performance, 20, 23, 29, 32

Motor, 177, 181, 182, 185, 198, 199, 204,

process, 23

206, 215, 219, 220, 226–229, 232–235

security, 306, 319

torque, 234
throughput, 299

Multi lingual hand writing recognition, 20

Neural

Multidimensional linear data, 61

network, 20, 23, 25, 52, 60, 62, 69, 74

Multi-language, 360
layers, 23

facility, 351
parameters, 20

software, 352
Neurons, 22, 23, 25, 31, 62

Multilayer
New revenue streams, 96

nonlinear transformation, 61
Nickel

perceptron, 21
cadmium, 157

Multi-level component analysis, 202

metal hydride, 157

Multilingual
Nitrogen, 189, 191, 194

computing, 351
Node, 23, 27, 70, 71, 73–76, 78, 111,

facility, 351
112, 120, 122, 123, 132, 133, 139, 140,

information management, 351

289–294, 296, 298, 302, 320

Multi-media division, 123

localization, 69–71, 77–79

Multiple
transmission range, 290

heterogeneous sensors, 37
Noise, 4, 37, 40, 220, 334, 335, 337

Index 411

Nominal management endeavor, 121

On-demand

Nonconventional energy, 238

network, 132

Non-digital platforms, 100

routing protocol, 291

Non-electronic, 169, 170, 172, 175,

One-way communication, 102

182–185
Online

design, 183
business frameworks, 91

device, 169, 170, 184

office software, 123

eye dropper, 183, 184

payment, 102

model, 170, 175, 182

software, 112

Non-horizontal line, 21
storage, 112, 123

Non-human understandable characters, 21

strategic policies, 89

Nonlinear controllers, 245

transaction, 102

Non-orthogonal, 199
Operating

Non-popularity, 172
model, 98

Non-readable text, 308

radio, 39

Non-repudiation, 276
temperature, 241

characteristics, 323
units, 105

Non-standard hieroglyphs, 276

voltage range, 178

Non-stationary signals, 197, 199

Opportunistic load balancing algorithm

Non-traditional methods, 134

(OLB), 112, 123, 135

Normal
Oppressive loaded system, 112

encryption algorithms, 287

OptiAid, 175

photo diodes, 178

Optic disc, 4, 10

recurrence of vibration, 364

Opticare eye drop dispenser, 174

Normalization
Optimal

algorithm, 25
energy, 254

graph cut segmentation scheme, 38

mother wavelet, 198, 199, 202–204, 215

Noteworthy speed, 83
parameter settings, 102

Novel corona virus, 48, 383

resource exploitation, 121

Nucleic acids, 322, 323

temperature, 389

Numerical
Optimization, 11, 19, 20, 22, 24, 27–32, 60,

analysis, 353
254, 334

data samples, 204

algorithm, 11, 24, 27–30, 32

Nutrient composition, 189

business model, 99

hyper parameter, 20

O values, 31

Object
Optimizer, 24, 28

classification, 60
Optimum

detection, 44
mother wavelet, 204, 205

Objective function, 22, 31

productivity, 188

Ocular
Organic

condition, 170, 171

framework, 90

disturbance, 170
matter, 189

drug, 169, 172, 184, 185

molecules, 384

surface, 171, 173, 177, 184

Organization, 20, 86, 88, 89, 93, 95–97, 291

Onboard
challenges, 97

computer, 36
Orthogonal wavelet, 199, 200

processing environment, 38
decomposition, 199

412 Index

Oscillating signal, 180

carcinogenesis, 385

Oscillations, 227
diodes, 178

Outdoor augmenting, 149

therapy, 392, 394

Output
transistors, 178

gate, 50, 52
Physical

layer, 22, 23, 27, 31, 69, 74

abused, 151

level, 178
machine, 110

Overlapping image patches, 13

games, 132

systems, 110

P workstations, 120

Pigment epithelium, 10

Packet loss, 299, 301

Pigmentary epithelial atrophy, 10

Pairwise co-prime positive integer, 321

Pitch

Pandemic, 48, 49, 52, 53, 56, 383

control, 254, 259–261, 266–271

Paper documents, 385

Parameter
speed, 257

optimization, 19, 22
Pixel, 9, 10, 25–27, 61, 62, 310

search space, 22
intensity levels, 6

specific learning rates, 24

Plain text, 277, 283, 324

Parasitic capacitance, 227

Plant monitoring, 188

Passive underlying vibration control, 377

system, 188

Path establishment, 302

Plastic hinge formation, 365

Pathological
Plateau region, 24

artifacts, 4
Police, 158, 159, 162, 163

samples, 10, 15, 16

Pollution, 220, 238

symptoms, 15
Pooling of sources, 110

Pay per use, 111

Popular transposition techniques, 324

Payment
Population density, 48, 49, 53

flexibility, 98
Portfolio management, 349, 350

patterns, 102
Position routing mechanism, 291

P-box permutation, 281

Positive

Peak signal noise ratio (PSNR), 3, 8, 12, 13,

big (PB), 246, 247

16
small (PS), 246, 247

Perfect forward secrecy, 285

Post-operative

Performance, 27, 122, 302, 344

cases, 171, 184, 185

analysis, 21, 334

drugs, 170

evaluation, 29, 292

Potassium, 188, 189, 191, 194

indices, 237
Potential customers, 100–102

management, 97
Pothole, 36

measurement, 15, 292

contours, 42, 43

index, 16
data function, 37

system (dynamic state), 229

detection, 37, 38, 41, 44

Permutations, 285
detector, 38, 41, 43

Personalization, 84, 92, 93

images, 38

guidance, 132
tracking system, 37

Petabytes, 86
Power

Phosphorous, 189, 191, 194

consumption, 185, 219

Photo
grid network, 255

aging, 385
output, 234

biological effect, 383, 394

quality, 256, 257, 265, 267, 270, 271

Index 413

regulators, 180, 185

ANN model development, 74

spectral density (PSD), 333–335, 337–344

model training and testing, 75

system, 253, 254, 257

system model, 38

Precautionary measures, 52
driver attention monitor, 39

Predicting, 47, 49, 50, 52, 53, 55, 56, 61, 69,
pothole detector, 40

70, 72, 76, 77

technique, 293

trends, 100
working steps (proposed work), 293

Preliminaries, 321
Protocols, 110, 183, 277, 278, 290, 291

asymmetric key cryptosystem, 323

Prototype wavelet function, 203

chinese remainder theorem, 321

Psychographic, 85

DNA encoding, 321

Public

image enhancement evaluation, 16

communication channels, 277

rail fence technique, 324

key, 281, 284, 286

stage, 3
cryptography, 278, 309, 323, 325, 327

XOR algebraic operation for DNA

encryption, 307

sequence, 322
utility accounting, 351

Pulsating output, 220

Pre-operative, 170

Punching resistance, 379

Pre-process, 37, 38

Purine

image datasets, 10

adenine, 322

input function, 63

guanine, 322

Pressure, 70, 171, 176, 389

Push-button switches, 156

lowering drugs, 170

PV array output voltage, 243

Pre-trained ResNet18, 62
P-V characteristics, 241, 242

Price
PV module, 238–243, 249

comparisons, 99
Pyrimidine

fluctuation, 101
cytosine, 322

optimization, 101
thymine, 322

services, 103
Python programming, 52

strategies, 101

Prime number, 279, 281, 284, 285

Q
Privacy enhancement technologies, 292

Private key cryptography, 308

Quadrature mirror filters, 201

Probabilities
Quake vibration, 377

density function, 6
Qualitative

distribution, 4
analysis, 60

of occurrence, 6
evaluation, 8

Process analysis, 97
Quantitative assessment, 173

Procurement methods, 103
Quantum computing, 105

Product performance, 102

Professional accounting terms, 358, 360

R
Proper
Rail fence technique, 320, 324, 327, 330

energy management, 250

Random

secured supervision, 157, 165

dynamic mapping, 310

technologies, 20
forest, 21

Proposed
overlapping patches, 12

method, 112, 135

variables, 6

algorithm, 136
Range

model, 74
algorithms, 71

414 Index

localization algorithms, 71
GPS-SMS child tracking system (smart

positioning methods, 71
phone), 147

Rapid flexibility, 110

IoT wearable smart health monitoring

Raspberry pi, 62
system, 148

platform, 62
pedestrian tracking (shoe-mounted iner­
Rate of transmission, 47, 53, 56
tial sensors), 149

Reaching behind, 39
smart IoT device (child safety-tracking),

Reactive
146

power, 257, 262, 264, 267, 269, 270

smart wearable bluetooth fitness tracker,

routing protocol, 291

148

Real
temperature regulating timepieces, 149

life dataset, 4
Renewable

time
energy

location, 145, 147, 148, 151, 157

power stations, 238

monitoring, 35
resources, 238

world
sources, 219, 220, 238, 254

applications, 61
power, 255

situations, 184
Rescue operation, 159, 160

Reasonable price, 98
Residual eye drops, 173

Received signal strength (RSS), 70

ResNet, 63–67

ResNet-110, 63

Receiver

ResNet-18, 62–65, 67

initiated, 121, 133

ResNet-50, 62–67

category algorithms, 133

model, 62

operator characteristic graph, 14

Resolution variable, 124

start up category methods, 121

Resource

Reconstructed

allocations, 120

image, 10, 62

parity, 292

program, 198, 199

pooling, 110

signal constituents, 202

utilization, 122

synthesis, 201
Respiratory suffocation, 392

waveform, 207–212
Respondents personal information, 353

Rectified linear unit (ReLU), 11, 27, 63, 64,

Response time, 122

74, 76
Results discussion, 40, 52, 183, 294, 337

Rectifier unit, 262

simulation design, 294

Recurrent neural network (RNN), 20, 50, 60

Retail

Redness-relieving drops, 170

business, 98, 99, 103

Reduced power losses, 234

industry, 96, 100–103, 105

Region, 6, 40, 43, 100, 101, 384

locations, 98

of interest (ROI), 40–43

market, 350, 360

Regional
sector, 96, 98, 103

language, 194, 349–353, 355–360

Retailer website, 102

facility, 349–352, 355–360

Retinal

transliteration, 360
artifacts, 7, 73

Regressive motion sensors, 60

fundus, 3, 5, 16

Regularity, 203
images, 3–5, 16

Regularization method, 25, 30, 31

image, 4, 5, 8–10, 12, 73

Reinforced concrete frame, 363, 379, 380

enhancement tool, 12

Related work, 85, 146, 334

surface, 4

Index 415

vessel segmentation dataset, 9

Scikit learn machine-learning analysis, 52

structures, 12
Scratched training, 61

vessel segmentation, 3–5, 16

Screw system, 169, 176

Revenue model, 97, 98

Season, 100, 101

Revolutionary contribution, 98
wise trends, 100

RGB color sensor, 190

Secondary data, 98

Ribonucleic acid (RNA), 384

Security, 92, 105, 123, 145, 146, 151, 162,

Road, 35, 36, 44, 88, 295

163, 165, 166, 273, 276, 286, 287, 291,

accidents, 35, 36
292, 305–309, 316, 319–321, 323, 330,

camera, 36
334

conditions, 35, 36, 40

devices, 330

side unit (RSU), 290, 293

measures, 105

surface, 36
protection, 292

sweeping vehicle automation, 291

sector improvement communication, 333

Roadside
stricken laws, 92

infrastructure, 293
Segmentation, 3–5, 9, 16, 21, 37, 38, 60,

units (rsus), 290, 293, 294, 296

102

Robotics, 44
accuracy, 16

Robust security system, 319

barriers, 102

Role of,
Seismic

big data in retail, 95, 98

dampers, 378, 379

retail business model implementing big

ground vibrations, 364

data, 99
load, 364–366, 376, 377, 379, 380

traditional retail business model, 98

performance, 378

Root mean squared error (RMSE), 69, 71,

tremors, 364

72, 74, 76, 78, 79, 198, 199, 203–212, 215

vibration, 377, 379

Rotor
waves, 364

position, 228
zone, 365, 379

speed, 226, 229, 234, 258–260

factor, 365, 376, 377

unbalance, 210
Seismicity, 364, 380

Route
Self-configuration, 289, 302

discovery process, 291

feature, 290

overhead, 299, 302

Sender

Routers, 290
beginner category, 121

Row-sum, 112
initiated, 121, 133

Rubber bearings, 377

Senior citizens, 101

Sensitivity, 9, 14–16, 102

S analysis, 102

Safe driving, 39
Sensor node, 69–78, 295

Sample data, 22, 204

SEPIC, 219, 221–223, 229, 231, 234, 235

Sanitization, 383
buck boost converter, 219, 234

Saturation current, 241

Sequence assemblage, 124

S-box, 281
Series-parallel resistances, 239

substitution, 280
Service provides on-demand, 110

Scalability, 110, 122, 134

Shear walls, 378

Scaling, 61, 91, 200

Shopper information, 85

School of Illumination Science Engineering

Short term fourier transform (STFT), 198–200

(SISED), 389
Shunt resistance, 241

416 Index

Sigmoid
Special

function, 51, 52
database 1, 25

output, 52
database 3, 25

Signal
Speed

compression, 197
control, 254

decomposition, 197, 200, 201

convergence to zero, 203

noising, 197
Spores, 389

processing, 199, 200

Sportspersons, 148

Simple trackers, 158

Spring frameworks, 377

Single
SPV array output power, 234

blinded technique, 183

Square root of the variance, 203

phase voltage unbalance, 204

Stacksparse coding, 61

Sinusoidal
Stakeholders, 349, 350, 359

excitations, 364
Standard

waveform, 262, 265, 270

deviation (SD), 72, 203

Skewness, 21
test conditions (STC), 238, 241–243

Slab-column connection, 379

State

Small
farm dataset, 39

business enterprises, 103

of-the-art, 20

key length, 286

Static

pocket sized device, 172

algorithms, 111, 122

Smart phone, 105, 147, 148, 188

compensator (STATCOM), 253, 254, 257,

camera, 38
260, 262, 264–271

Smooth power generation, 260

infrastructure, 290

Social
load balancing algorithms, 111

distancing, 48
system algorithm, 121

media, 100
Stationary signal analysis, 198

websites, 101
Statistical

Socio-economic, 85, 92
analysis, 56, 97, 202

benefits, 105
error

problems, 238
IQA metric, 8

Sodium hypo-chloride, 383
metrics, 8, 16

Soil surface, 36
parameters, 202

Solar, 219, 220, 222, 228, 229, 231, 234,

Stator winding fault, 204

238, 239, 254

Steganography, 305–308, 316

energy, 238
Stego image, 307, 316

irradiation
Sterilization

availability, 219
indoor areas, 384

levels, 234
temperature, 389

panel, 220
Stochastic

power generation technology, 220

approximation, 24

Sophisticated algorithms, 105

gradient descent (SGD), 11, 24, 30, 31

SOS button, 146, 158, 159, 161–166

momentum gradient descent (SGDM), 63

Sound
Storage, 20, 109, 110, 131, 132, 220, 320

component, 185
Store extension, 88

retail business system, 98

Storey

Space studies, 44
displacement, 367, 369–375, 379

Sparse interpretation, 61
drift, 367, 370–372, 374, 375, 379

Index 417

Story drift, 364, 367, 370, 371, 373, 374,

network model, 71

376–380
useful definitions, 71

Strategic
stability, 255

decision-making, 105
topology, 290

marketing decision, 102

Systematic

planning, 100
analysis, 61

Strong visual models, 60

cost model, 98

Structural
Systemic illness, 176

elements, 8, 73, 378

integrity, 10
T
Substantial growth, 254
Tablet, 84

Support vector machine (SVM), 21

Tabular validation, 53

Sustainable economic development, 103

Tanh function, 51, 52

Switches, 156
Target

pulse frequency, 225

marketing, 103

Sybil, 290, 292, 293, 296, 300, 302

revenue, 98

attack, 290, 292, 293, 296, 300, 302

Tear duct, 171

vanets, 292
Technique comparison, 299, 300

Symlet5 mother wavelet, 215

Technological

Symlets (Sym), 199, 200, 204, 207

advancement, 103

Symmetric, 121, 133, 307, 308, 320

availability, 188

algorithms, 285
penetration, 104

encryption, 307
Telecommunications, 96

key
company, 96

algorithm, 285, 323

Temporary fertility, 187

encryption algorithms, 287

Tensor flow, 12, 27, 37, 39, 62

encryption, 320
Tenterhooks, 161

Symmetry, 203
Terabytes, 86

System
Test

communications, 132
accuracy, 14, 16, 22, 26, 27

congruence, 321
input matrix, 75, 76

sensor node, 76

design, 190, 191, 222, 335

Text steganography, 308

block diagram, 191

Textiles materials, 385

calculation of voltage current ripples

Texting, 39

(convertor), 225

Theoretical background, 22

circuit diagram, 191

classifier optimization, 22

control of proposed system, 227

deep neural network (DNN), 22

DC link capacitance of VSI, 226

network hyper parameters, 23

DC-DC convertor, 222

batch size, 25

hardware required, 190

dropout, 25

irradiation levels, 229

hidden layers, 24

P and O MPPT algorithm, 227

learning rate, 24

PV panel, 222
optimizer, 24

software required, 190

network layers, 23

water pump, 227

hidden layer, 23

model, 39, 40, 71

input layer, 23

morphological operation enhancement,

output layer, 23

73
Thermal voltage, 241

418 Index

Three-layered architecture, 97
True

Thresholding schemes, 199

negative (TN), 9, 26, 63

Throat cavity, 171

positive (TP), 9, 26, 63

Throughput, 121–123, 293, 297–300, 302

Tubular eyedropper, 174

Thymine, 322
Tunable isolator, 377

Time
Turbine generator, 267

complexity, 22, 32
set, 265

consuming, 5, 20, 24
Turbulence, 257

process, 187
Twitter notification, 148

difference of arrival (TDOA), 69–72, 78, 79

Two-dimensional signals, 306

duration, 198, 229, 234

Two-phase scheduling, 135

frame, 98

Tophat bottom-hat transformations, 7, 73

U
Topical
Ultraviolet, 384, 394

drugs, 170, 171

ray, 384

hypotensive medication, 173

Unassigned jobs, 136, 138

therapy, 170
Unauthorized

Tracking
access, 276, 319

devices, 149
user, 276, 307, 309

problem, 151
Unbalanced

Traditional
cost matrix, 112

buck-boost converters, 220

loads, 120

business data storage, 110

matrix, 112, 117, 135, 139

cryptography system, 320

resource usage, 120

cryptosystems, 321
rotor, 204

instillation, 173
Uncommitted component analysis, 334

retail business, 99
Unencrypted data, 277

model, 102
U-net

voltage control techniques, 262

model, 10

working, 257
type connections, 10

Traffic
Uneven illumination, 4

counting, 60
Union territories, 49

rules, 36
Unpredictable workloads, 120

scenario, 60
Unreadable format, 307

Trafficked, 151
Upsampling, 201

Training, 10, 12, 31, 64–66

Urban business area, 160

details, 5
Used cases, 160

Transducer, 180
child safety, 162

Transfer learning, 39, 40, 61, 62

fire fighter safety, 160

Transient voltages-currents, 257

fishermen safety, 161

Transmission
women safety, 162

lines, 267, 271

UV radiation, 385, 387

process, 319, 320

system handling operators (TSO), 255

V

Transposition technique, 324

V2I communications, 292

Trapdoor function, 285

Validation, 12, 14, 22, 26, 27, 30, 31, 53, 55,

Trend
62, 64–66, 97

forecasting algorithms, 101

accuracy, 26

identification, 100
phase, 26

Index 419

set, 26
ripple, 226

Value information, 86
sag, 257

Vanishing moments number, 203

swell, 257, 269, 271

Variable
switching inverters (VSI), 219–222,

pitch technology, 254

226–228, 234

rotation speed, 254

Volt-second balance equation, 223, 225

speed, 255, 257, 260, 264, 268, 271

Vascular abnormalities, 10
W
Vehicle
Walmart, 86–88, 93

ad-hoc networks, 290

labs examinations, 86

automation, 291
stores, 86

communication, 291
Wastage of,

to vehicle (V2V), 290, 292, 295

drug, 171–173, 183, 184

traffic density, 291

quantity, 170, 172

Vernacular language, 352

medicine, 172

Versatile

Water pumping

base separation framework, 377

applications, 219, 220

infrastructure, 289

system, 219, 221, 234

Vertical acceleration, 364

Wavelet, 197–200, 203, 204, 206

Vessel

analysis, 199, 203

classification, 4, 5, 9, 14, 16

coefficients, 197–200, 206

identification, 10

decomposition, 200, 201

probability map, 5

segmentation, 3–5, 16, 17

family, 200

Vibratory style, 364

functions, 199, 200, 206–212

Video
transform, 197, 199, 200, 204, 205

frame, 306
Wearable

processing, 38
device, 151, 157

spilling applications, 92
tracker, 146

steganography, 308
Weather, 36, 161, 164

Virtual
Web

machine, 119–121, 132

browsing history, 101

monitor, 121, 132

business, 85, 91

Virtualization
field, 85

concept, 121
life information, 86

techniques, 132
showcasing systems, 83

Viruses, 384, 389, 392

Websites replacing branch agencies, 110

Visual
Weighting distribution (WD), 6, 7

clues, 38
Wind

impairment, 169, 183, 184

energy

information
assimilation, 253

IQA metric, 8
technology, 254

metrics, 8
generation, 253, 255

perception IQA metrics, 16

installation capacity, 253

Vital body parameters, 146, 149, 165

generator, 254, 255

Voltage, 180, 205, 206, 212, 219–222, 226,

outturns, 257

228–230, 239–244, 246, 248–250, 254,

intermittency, 254

257, 262, 268–271

power, 254, 255, 260

gain, 221
power in India, 255

420 Index

integration of wind power (existing

World
grid network), 256
carbon footprint, 254

proportional control pitch actuator

electricity consumption, 254

system (CSA), 257

Health Organization (WHO), 48, 50

shear, 257
Wormhole attack, 290

turbine, 254–257, 260, 264, 268

Wireless
X

connection, 148
XOR operation, 322, 323

links, 70

sensor network (WSN), 69–71, 78, 79

Z
Workstation, 110, 111, 120, 121, 132, 133

consolidation, 120
Zero padding, 201

utilization, 120
Zigzag cipher, 324