Biometrics: Theory, Methods, and Applications

Book Preface

The objective of biometric systems is the recognition or authentication of individuals based on some physical or behavioral characteristics that are intrinsically unique for each individual. Nowadays, biometric systems are fundamental components of advanced security architectures. The applications of biometrics range from access control, military, and surveillance to banking and multimedia copyright protection. Recently, biometric information has started to become an essential element in government issued authentication and travel documents. The large-scale deployment of biometrics sensors in a variety of electronic devices, such as mobile phones, laptops, and personal digital assistants (PDA), has further accelerated the pace at which the demand for biometric technologies has been growing. The immense interest in the theory, technology, applications, and social implications of biometric systems has created an imperative need for the systematic study of the use of biometrics in security and surveillance infrastructures.

This edited volume provides an extensive survey of biometrics theory, methods, and applications, making it a good source of information for researchers, security experts, policy makers, engineers, and graduate students. The volume consists of 26 chapters which cover most aspects of biometric systems. The first few chapters address particular recognition techniques that can be used in conjunction with a variety of biometric traits. The following chapters present technologies tailored to specific biometric traits, such as face, hand geometry, fingerprints, signature, electrocardiogram, electroencephalogram, and gait. The remaining chapters focus on both theoretical issues as well as issues related to the emerging area of privacy-enhancing biometric solutions.

An overview of recent developments in discriminant analysis for dimensionality reduction is presented in the first chapter. Specifically, a unified framework is presented for generalized linear discriminant analysis (LDA) via a transfer function. It is shown that various LDA-based algorithms differ in their transfer functions. This framework explains the properties of various algorithms and their relationship. Furthermore, the theoretical properties of various algorithms and their relationship are also presented. An emerging extension of the classical LDA is the multilinear discriminant analysis (MLDA) for biometric signal recognition. Biometric signals are mostly multidimensional objects, known as tensors. Recently, there has been a growing interest in MLDA solutions. In Chapter 2, the fundamentals of existing MLDA solutions are presented and then categorized according to the multilinear projection employed. At the same time, their connections with traditional linear solutions are pointed out. The next two chapters present classification issues in biometric identification. The problem of classification is extremely important because it essentially sets the framework regarding the way decisions are made once feature extraction and dimensionality reduction have taken place. A variety of classification approaches can be taken. One of these approaches is to use neural networks (NN). Chapter 3 is a comparative survey on biometric identity authentication techniques based on neural networks. This chapter presents a survey on representative NN-based methodologies for biometric identification. In particular, it captures the evolution of some of the representative NN based methods in order to provide an outline of the application of neural nets in biometric systems. A specific, but far from uncommon, case of classification is that involving fusion of biometrics. The main task here is the design of classifiers for fusion based biometric verification, which is addressed in Chapter 4. The chapter provides guidelines for optimal ensemble generation, where each classifier in the ensemble is a base classifier. Examples are shown for support vector machines and correlation filters. The chapter also focuses on decision fusion rules and the effect of classifier output diversity on their decision fusion accuracy is also analyzed.