Machine Learning For Biometrics: Concepts, Algorithms and Applications (Cognitive Data Science in Sustainable Computing) 1St Edition - Ebook PDF
Machine Learning For Biometrics: Concepts, Algorithms and Applications (Cognitive Data Science in Sustainable Computing) 1St Edition - Ebook PDF
com
https://ebooksecure.com/download/introduction-to-algorithms-for-data-
mining-and-machine-learning-ebook-pdf/
ebooksecure.com
https://ebooksecure.com/product/ebook-pdf-machine-learning-refined-
foundations-algorithms-and-applications/
ebooksecure.com
https://ebooksecure.com/product/ebook-pdf-microeconomics-theory-
applications-with-calculus-4th/
ebooksecure.com
Hands-On Ethical Hacking and Network Defense (MindTap
Course List) [Team-IRA] [True PDF] 4th Edition Rob Wilson
- eBook PDF
https://ebooksecure.com/download/hands-on-ethical-hacking-and-network-
defense-mindtap-course-list-team-ira-true-pdf-ebook-pdf/
ebooksecure.com
https://ebooksecure.com/product/ebook-pdf-digital-systems-engineering/
ebooksecure.com
https://ebooksecure.com/product/ebook-pdf-organizational-behaviour-
concepts-controversies-7th/
ebooksecure.com
https://ebooksecure.com/product/ebook-pdf-consumer-behaviour-2nd-
edition/
ebooksecure.com
https://ebooksecure.com/product/ebook-pdf-listening-to-western-
music-7th-edition-by-wright/
ebooksecure.com
(eBook PDF) Business Law 6th Edition by James Marson
https://ebooksecure.com/product/ebook-pdf-business-law-6th-edition-by-
james-marson/
ebooksecure.com
Machine Learning for Biometrics
Concepts, Algorithms, and Applications
Cognitive Data Science in Sustainable Computing
Edited by
Madhumita Panda
Assistant Professor, Master in Computer Applications, Seemanta Engineering College,
Jharpokharia, Baripada, Odisha, India
Subhashree Mishra
Assistant Professor, School of Electronics Engineering, KIIT University, Bhubaneswar,
Odisha, India
Banshidhar Majhi
Director, IIITDM, Kancheepuram, Chennai, Tamil Nadu, India
Series Editor
Arun Kumar Sangaiah
School of Computing Science and Engineering,Vellore Institute of Technology (VIT),
Vellore, India
Academic Press is an imprint of Elsevier
125 London Wall, London EC2Y 5AS, United Kingdom
525 B Street, Suite 1650, San Diego, CA 92101, United States
50 Hampshire Street, 5th Floor, Cambridge, MA 02139, United States
The Boulevard, Langford Lane, Kidlington, Oxford OX5 1GB, United Kingdom
Copyright © 2022 Elsevier Inc. All rights reserved
No part of this publication may be reproduced or transmitted in any form or by any means,
electronic or mechanical, including photocopying, recording, or any information storage and
retrieval system, without permission in writing from the publisher. Details on how to seek
permission, further information about the Publisher’s permissions policies and our arrangements
with organizations such as the Copyright Clearance Center and the Copyright Licensing Agency, can
be found at our website: www.elsevier.com/permissions.
This book and the individual contributions contained in it are protected under copyright by the
Publisher (other than as may be noted herein).
Notices
Knowledge and best practice in this field are constantly changing. As new research and experience
broaden our understanding, changes in research methods, professional practices, or medical
treatment may become necessary.
Practitioners and researchers must always rely on their own experience and knowledge in evaluating
and using any information, methods, compounds, or experiments described herein. In using such
information or methods they should be mindful of their own safety and the safety of others,
including parties for whom they have a professional responsibility.
To the fullest extent of the law, neither the Publisher nor the authors, contributors, or editors,
assume any liability for any injury and/or damage to persons or property as a matter of products
liability, negligence or otherwise, or from any use or operation of any methods, products,
instructions, or ideas contained in the material herein.
Numbers in paraentheses indicate the pages on which the authors’ contributions begin.
Sumitav Acharya (143), Department of Computer Science and Engineering, National
Institute of Science and Technology, Berhampur, India
A. Anandh (105), Department of Computer Science and Engineering, Kamaraj College
of Engineering and Technology, Madurai, India
Saurabh Bilgaiyan (155), School of Computer Engineering, KIIT Deemed to be
University, Bhubaneswar, Odisha, India
P.V.S.S.R. Chandra Mouli (65), Department of Computer Science, Central University
of Tamil Nadu, Thiruvarur, India
Manisha P. Dale (1), MES College of Engineering, Savitribai Phule Pune University,
Pune, India
Rupam Das (155), School of Electronics Engineering, KIIT, Deemed to be University,
Bhubaneswar, India
K. Devendran (87), Computer Science and Engineering, Kongu Engineering College,
Erode, Tamil Nadu, India
Sachit Dhamija (155), School of Electronics Engineering, KIIT, Deemed to be
University, Bhubaneswar, India
R. Jai Ganesh (129), K. Ramakrishnan College of Technology, Trichy, India
G.K. Kamalam (177), Department of Information Technology, Kongu Engineering
College, Erode, Tamil Nadu, India
Vaishali H. Kamble (1), AISSMS IOIT, Savitribai Phule Pune University, Pune, India
M. Kavitha (129), K. Ramakrishnan College of Technology, Trichy, India
P. Keerthika (87,177), Department of Computer Science and Engineering, Kongu
Engineering College, Erode, Tamil Nadu, India
Chirag Kyal (29), Department of Computer Science and Engineering, National Institute
of Science and Technology, Berhampur, India
K. Logeswaran (177), Department of Information Technology, Kongu Engineering
College, Erode, Tamil Nadu, India
V.M. Manikandan (201), Computer Science and Engineering, SRM University-AP,
Amaravati, Andhra Pradesh, India
R. Manjula Devi (87,177), Department of Computer Science and Engineering, Kongu
Engineering College, Erode, Tamil Nadu, India
xiii
xiv Contributors
xv
xvi Preface
The editors would like to gratefully acknowledge all of the contributors for con-
tinuous effort and timely submission of their chapters. This book would not
have been feasible without the cooperation of the chapter authors. All the chap-
ters have been reviewed for several rounds to facilitate the selection of final
chapters in our book. Valuable suggestions and guidance from the reviewers
helped the authors in refining individual chapters. Thanks also go to the
reviewers in enhancing the quality of the chapters of this book.
The editors would like to extend gratitude to Sonnini Ruiz Yura from
Elsevier for inspiration over the year. We would like to gratefully acknowledge
Andrae Akeh and Judith Clarisse Punzalan (Elsevier) for their patience during
the preparation of this book. In addition, we shall thank Swapna Srinivasan at
Elsevier for her sincere help and patience during the final preparation of this
book. Finally, we shall extend gratefulness to our family members and friends
for all their support.
xix
Chapter 1
1 Introduction
Face of an individual is a popular and well-accepted biometric trait that can be
used to perform identity recognition of adults as well as children. Children are
the most valuable and jeopardy group in society; hence they should be under
supervision continuously. Security and healthcare of children is an important
aspect of all countries [1]. Automatic recognition of children using their face
is a useful investigative tool to help identify missing children. Though the
development of the face of a child starts in the mother’s womb from 3 months,
it is not proportional to the development of other parts of the body. Therefore,
recognition of children below 6 years is still an open research problem. Children
recognition using different modalities needs to be studied to solve the problems
related to security and healthcare. As per the literature survey, still there is not
even one commercial biometric system in use for recognition of toddlers. Var-
ious researchers have discussed about biometric recognition of adults; however,
very few papers on toddlers are available. A meager amount of work has been
traced in biometric identification of toddlers or children. It is most challenging
to recognize a toddler from his own single photograph after a few months. But,
in some instances, such as missing children, we have only the face image. So,
the recognition of children from their face image is very important. Facial
images can be acquired without users’ active involvement using ordinary cam-
eras from a distance. Also in survey, we noted that that most of the toddler’s
biometric recognition is in verification mode.
Sahar Siddique studied longitudinal face recognition using the Extended
Newborn Database and Children Multimodal Biometric Database. Identifica-
tion accuracy achieved using CNN is 62.7% and 85.1% on both the databases,
respectively [2]. Rowden et al. studied the longitudinal face recognition of chil-
dren between 0 and 4 years of age. The same-session accuracy they achieved is
93% and cross-session accuracy 43% after 6 months [3]. Local Binary Pattern
(LBP) is widely used texture-based method for recognition of face biometrics
[4]. This method is also used for children face recognition [5,6].
This chapter is presenting a machine learning and deep learning approach
for face recognition of children. As per the literature review, the study of chil-
dren recognition with the help of their face modality started in 2010 in India
[5,6]. Most of the papers studied same-session face recognition. The meager
amount of work is carried out longitudinally, that is, images of the same subject
taken over a period of time. There are very few readily available databases for
newborns like FG-NET [7]. There are very few images of children below
5 years in these datasets. CMBD database of IIIT Delhi (India) [8] is one of
the children longitudinal face recognition databases, but due to security reasons
the face database is not publicly available [9]. Therefore, database collection is
the major task in infants and toddler’s recognition. For this study, the database
of 81 subjects for the same session and 48 subjects for the cross session is col-
lected. The span between the data acquisition sessions is 3–6 months. The major
stages of the proposed work are preprocessing of face images, which include
manually cropping of face images of size 120 120 and converting it into gray
scale. Feature extraction is based on principal component analysis and linear
discriminant analysis, and CNN is proposed. The classification of subjects is
done using machine learning classifiers on the children database of the same
session and cross session. Further, the work is extended using convolutional
neural network (CNN) in which we have proposed our own optimized model
with data augmentation used to compare the machine learning and deep learn-
ing classifiers on our database.
The major contributions of the proposed work are as follows.
1. Due to the nonavailability of reference databases for children below 6 years,
the collection of longitudinal databases of children faces itself is a great
challenge. In the proposed work, we have captured the face images of tod-
dlers with the mobile camera of resolution 20 MP, with consent from their
parents in two sessions. Time period between two sessions is 2 months to
1 year. In the same session, 730 images of 81 toddlers are taken. In the sec-
ond session, 485 images of 48 toddlers are captured.
2. In this chapter, we are proposing feature extraction using PCA, LDA, and
CNN approaches on the facial images and their comparative study. In PCA
and LDA approaches, we have applied different machine learning tech-
niques for classification of subjects, such as Support Vector Machine
(SVM), Logistic Regression (LR), Gaussian Naive Bayes, K-Nearest
Neighbors (KNN), Decision Tree, and Random Forest. In CNN, feature
extraction is done by convolution layers and classification is done by dense
layers.
Machine learning approach Chapter 1 3
54.8
27.8 80.5
307.4
Fingerprint Recognition
Facial Authentication
IRIS Recognition
Voice Recognition
Palm Recognition
Others
448.2
1679.4
useful in the case of missing children. Therefore, we are using face biometrics
for recognition of children. An example of different biometrics of children from
our database is shown in Fig. 2.
Human face is a 3D model, we can recognize it by its features such as eyes,
shape of face, and color. Automatic face recognition is based on 2D photo-
graphs of a person. In 1964 and 1965, Woody Bledsoe, along with Helen Chan
and Charles Bisson, recognized human faces using computers for the first time.
Nowadays adult face recognition achieves very high accuracy in the range of
99.63% [11]. Face recognition of children is still an open challenge. All face
recognition algorithms evaluate on false negative identification rate (FNIR)
which is dependent on age. The FNIR of children is higher as compared to
adults or seniors. The comparison of FNIR and false positive identification rate
(FPIR) is given in Table 1 for various stages of child age.
Child face development: Studies from the paper by Farkas discuss that in
the first 2 years of a child face development, mouth width of child increases
whereas mouth height decreases. Mouth shape alters from “rosebud-like” to
a more adult type. Growth of facial features is very fast in the first year, less
rapid in the second year. Subsequent changes were slow and irregular from
the age of 3 to 9 years [12].