This book (consisting of 11 chapters) covers the state-of-the-art research on signal and image processing techniques used to support biometric recognition. There are many real-world applications, such as surveillance and identity verification, which require the development of a pattern recognition process.

Chapter 1 presents the concepts of manifold learning and information geometry, and then discusses how manifold geometry can be exploited to obtain biometric data representations. Manifold learning is an important trend in pattern recognition methods for some data and information dimensionality issues.

Chapter 2 proposes some issues in remote face recognition, where faces are several tens of meters from the cameras. This task is substantially more challenging than constrained acquisition conditions. It has been demonstrated that finding features and statistical models, which account for the variations of the environment conditions, are the critical factors for face recognition performance.

Chapter 3 compares human and machine performance in the task of sketch identification. The analysis shows that the sparse representation for classification algorithm and Gabor-based description yields the best results.

Chapter 4 focuses on recognizing faces with variations in aging and disguise. The performance of the proposed algorithm is demonstrated on databases that cover a comprehensive set of aging and disguise scenarios.

Chapter 5 reviews three face recognition algorithms and compares the fusion performance of seven fusion methods. The author obtains promising results on a multispectral face dataset. The experimental results show that the k-nearest neighbor (KNN) score fusion gives the best overall performance and that all of the evaluated score fusions can increase accuracy.

Chapter 6 presents a survey of current techniques for ear recognition, from geometrical to 2D-3D multimodal. It discusses some open problems, such as ear detection, illumination, pose variations, and outdoor conditions.

Chapter 7 introduces a “quality-based unconstrained eye recognition” system that uses both sclera and iris recognition for human identification. Iris recognition only works for near infrared images while sclera recognition works reasonably well with visible wavelengths, so the combination can be advantageous.

Chapter 8 proposes a method of speed-invariant gait recognition in a unification of model- and appearance-based approaches. The proposed algorithm uses static features rather than kinematic features. The evaluation shows its advantage in nonconstrained acquisition scenarios.

Chapter 10 tackles the problem of quality criteria for online handwritten signatures, namely signature complexity and signature stability. The authors analyze such criteria with a unifying view in terms of entropy-based measures. The experiments show that the degradation of signatures due to mobile acquisition conditions can be quantified by entropy-based measures.

Chapter 11 starts with an overview of modern biometric authentication systems. The authors focus on the tracking capabilities of the acquisition process. The fundamental problem can be broken down into three issues: the vision system, software, and hardware, which are closely related to each other.

This book offers a good introduction to junior graduate students who are new to the field, and is also a good comprehensive reference for experienced practitioners. It covers a large variety of biometric traits, such as face, fingerprint, iris, gait, and signature. Readers should refer to machine learning books for more specific algorithms, to get an insightful understanding, because some techniques discussed in this book are only used as applications but are not illustrated in detail.