Table Of ContentJames Eric Mason · Issa Traoré
Isaac Woungang
Machine Learning
Techniques for
Gait Biometric
Recognition
Using the Ground Reaction Force
Machine Learning Techniques for Gait Biometric
Recognition
é
James Eric Mason Issa Traor
(cid:129)
Isaac Woungang
Machine Learning
Techniques for Gait
Biometric Recognition
Using the Ground Reaction Force
123
James EricMason IsaacWoungang
University of Victoria Ryerson University
Victoria, BC Toronto, ON
Canada Canada
Issa Traoré
University of Victoria
Victoria, BC
Canada
ISBN978-3-319-29086-7 ISBN978-3-319-29088-1 (eBook)
DOI 10.1007/978-3-319-29088-1
LibraryofCongressControlNumber:2015960234
©SpringerInternationalPublishingSwitzerland2016
Thisworkissubjecttocopyright.AllrightsarereservedbythePublisher,whetherthewholeorpart
of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations,
recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission
orinformationstorageandretrieval,electronicadaptation,computersoftware,orbysimilarordissimilar
methodologynowknownorhereafterdeveloped.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publicationdoesnotimply,evenintheabsenceofaspecificstatement,thatsuchnamesareexemptfrom
therelevantprotectivelawsandregulationsandthereforefreeforgeneraluse.
The publisher, the authors and the editors are safe to assume that the advice and information in this
book are believed to be true and accurate at the date of publication. Neither the publisher nor the
authorsortheeditorsgiveawarranty,expressorimplied,withrespecttothematerialcontainedhereinor
foranyerrorsoromissionsthatmayhavebeenmade.
Printedonacid-freepaper
ThisSpringerimprintispublishedbySpringerNature
TheregisteredcompanyisSpringerInternationalPublishingAGSwitzerland
To mom, dad, and my wife Pairin, you
helped me get to where I am today. Sombat
and Kangwon, oath fulfilled.
To Me-Kon, Mustapha, Khadijah, Kaden,
and Ayesha, for your unconditional love
and making me a blessed and lucky man.
To Clarisse, Clyde, Lenny, and Kylian,
for being there for me all the time. Your
endless love, support and encouragement,
and push for tenacity, are very much
appreciated.
Preface
Thelasttwodecadeshaveseenadramaticincreaseinthenumberofstakeholdersof
biometric technologies. The quality of the technologies has increased due to an
improvement in underlying data processing and sensor technologies. A growing
andhealthymarketplacehasemerged,whilethenumberofpeopleusing,operating,
or impacted by these technologies has been growing exponentially. Several new
disruptivetechnologieshaveemerged,alongwiththediversificationofthedevices
andplatformswherebiometricsareprovisioned.Theubiquityofmobilephonesand
the multiplicity and diversity of sensors available for biometric provisioning (e.g.,
webcam, fingerprint reader, touchscreen, accelerometer, gyroscope, etc.) is con-
tributing significantly to this dramatic growth of the biometric ecosystem.
Gaitbiometricsisoneofthenewtechnologiesthathaveappearedinthepastfew
decades.Gaitbiometrictechnologyconsistsofextractingandmeasuringuniqueand
distinctive patterns from human locomotion. Different forms of gait biometrics are
available based on how the gait information is captured (e.g., video cameras, floor
sensor, smartphones, etc.). Gait based on the ground reaction force (GRF) is the
most recent form of gait biometric technology, which although lesser known than
its counterparts, has shown greater promise in terms of its robustness. GRF is a
measure of the force exerted by the ground back on the foot during a footstep.
The GRF-based gait biometric is the central topic of this book. Theoretical and
practicalunderpinningsoftheGRF-basedgaitbiometricarepresentedindetail.The
main components and processes involved in developing a GRF-based recognition
systemarediscussedfromatheoreticalandexperimentalperspective,byrevisiting
existing research and introducing new results.
While the central topic of the book is GRF-based gait biometric technology, its
backdrop is machine learning. Several machine learning techniques used in the
literature for GRF recognition are dissected, contrasted, and investigated
experimentally.
The book covers the different dimensions required for developing a GRF-based
system:theoreticalmodels,experimentalmodels,andimplementationissues.Italso
coversindetailseveralmachinelearningalgorithmswhichcanbeusedbroadlyfor
vii
viii Preface
biometric recognition technologies and other similar pattern recognition problems
(e.g., speech recognition). This book is intended for researchers, developers, and
managers and for students of computer science and engineering, in particular
graduate students at the Master’s and Ph.D. levels, working or with interest in the
aforementioned areas.
The book consists of 11 chapters outlined as follows.
Chapter 1 provides a brief introduction to gait biometrics and outlines the
context, objectives, and main contributions of the book.
Chapter 2 gives a high-level overview of machine learning and presents the
different factors that define gait biometric technology. A high-level discussion
of the considerations and issues underlying the design of a gait-based biometric
recognition system is conducted.
Chapter 3 describes the field of gait biometrics and provides a historical over-
viewofworkthathasbeendoneinthefieldtodate.Itgoesontoexplainwherethe
footstep GRF fits into the field of gait biometrics, and reviews the footstep GRF
recognition literature that forms the basis for the research presented in later
chapters.
This chapter also presents the experimental setup and introduces the method-
ology used to achieve the research objectives. It covers the selection of a devel-
opment dataset containing data of a single shoe type and proposes a biometric
system composed of feature extractors, normalizers, and classifiers to perform
GRF-based person recognition.
Chapter 4 outlines the purpose and mechanisms underlying feature extraction,
and compares four different feature extraction techniques previously used for
GRF-based recognition in other studies. Theoretical background is provided for
each feature extractor together with a discussion of each implementation.
Preliminary GRF recognition results are acquired using the development dataset
and presented for the parameter optimization of each extractor.
Chapter5demonstratestheperformanceofvariousnormalizationtechniqueson
theextractedfeaturespacesfromChap.4.Twonovelnormalizationtechniquesare
introduced here and theoretical background is provided for these and several other
well-known existing techniques that are also examined. To determine the effec-
tiveness of normalization the results from applying these normalization techniques
are compared with the non-normalized results of the previous chapter.
Chapter 6 presents the theoretical background and implementation for five dif-
ferentclassifiersthatwereselectedforanalysisinthisbook.Eachclassifieristuned
across the best-performing feature spaces acquired from development dataset in
Chaps. 4 and 5. Finally, the feature extractor-normalizer-classifier combinations
that achieved the best results are summarized for comparison with the results over
the evaluation dataset in Chap. 8.
Chapter 7 outlines the experimental method and dataset used to evaluate the
proposed GRF-based gait biometric framework. Different evaluation criteria are
presented and discussed.
Preface ix
Chapter 8 demonstrates the results obtained after applying the best footstep
GRF-recognition systems, outlined in Chap. 6, to an evaluation dataset containing
previously unseen data samples with different shoe types.
Chapter 9 discusses the findings behind the GRF footstep recognition experi-
ment. The effects of various techniques are compared, with practical implications
and explanations for possible sources of error presented. Finally, the chapter con-
cludesbyexaminingtechniquesthatcouldpotentiallybeusedtoimproveuponthe
results discovered in our research.
Chapter 10 identifies various applications for gait biometrics, and discusses
current usage of the biometric both commercially and in research.
Chapter 11 provides a final summary of the research presented in previous
chapters.Themajorfindingsarehighlightedandtheremainingproblemsandareas
for future work are discussed.
The above chapters present state-of-the-art GRF-based gait biometric research
and expose different dimensions of this emerging field, by emphasizing both the
theoretical and practical underpinnings. We hope that this will represent a useful
resource for readers from academia and industry in pursuing future research and
developing new applications.
Acknowledgments
Asauthorsofthisbook,wewanttoacknowledgethosewhohelpedinitscreation.
In particular, we would like to thank Mr. Quin Sandler and Norman Mckay,
CEO and Product Architect at Plantiga Technologies Inc., for the fruitful discus-
sions on the industrial perspectives of gait biometric technologies, and helping in
our search for useful datasets.
Special thanks should go to the University of Calgary Faculty of Kinesiology
laboratory for sharing their GRF datasets.
Theresearchpresentedinthisworkwasmadepossiblethankstoresearchgrants
provided by the Natural Sciences and Engineering Research Council (NSERC) of
Canada.
xi
Contents
1 Introduction to Gait Biometrics . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Context. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Objectives. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Summary of Contributions. . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3.1 Feature Extraction . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3.2 Normalization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3.3 Classification. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3.4 Shoe Variation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2 Gait Biometric Recognition. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.1 Introduction to Machine Learning. . . . . . . . . . . . . . . . . . . . . . 9
2.1.1 Machine Learning Paradigm . . . . . . . . . . . . . . . . . . . 9
2.1.2 Machine Learning Design Cycle. . . . . . . . . . . . . . . . . 10
2.2 General Principles of Designing Gait Biometric-Based
Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.3 Authentication Using the Gait Biometric. . . . . . . . . . . . . . . . . 22
2.3.1 Privacy and Security Implications
of Gait Biometrics . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.3.2 Gait Biometric Approaches . . . . . . . . . . . . . . . . . . . . 24
2.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3 Gait Biometric Recognition Using the Footstep Ground Reaction
Force. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.1 The Ground Reaction Force. . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.2 Feature Extraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.3 Normalization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.4 Classification Approaches . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
3.5 Shoe Type. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
xiii
