Li Q. Speaker Authentication

Springer, 2012. — 264 p.This book is organized by research topic. Each chapter focuses on a major topic and can be read independently. Each chapter contains advanced algorithms along with real speech examples and evaluation results to validate the usefulness of the selected topics. Special attention has been given to the topics related to improving overall system robustness and performance, such as robust endpoint detection, fast discriminative training theory and algorithms, detection-based decoding, and sequential authentication. I have also given attention to those novel approaches that may lead to new research directions, such as a recently developed auditory transform (AT) to replace the fast Fourier transform (FFT) and auditory-based feature extraction algorithms. For real applications, a good speaker authentication system must first have an acceptable authentication accuracy and then be robust to background noise, channel distortion, and speaker variability. A number of speaker authentication systems can be designed based on the methods and techniques presented in this book. A particular system can be designed to meet required specifications by selecting an authentication method or combining several authentication and decision methods introduced in the book.
Speaker authentication is a subject that relies on the research efforts of many different fields, including, but not limited to, physics, acoustics, psychology, physiology, hearing, auditory nerve, brain, auditory perception, parametric and nonparametric statistics, signal processing, pattern recognition, acoustic phonetics, linguistics, natural language processing, linear and nonlinear programming, optimization, communications, etc. This book only covers a subset of these topics. Due to my limited time and experience, this book only focuses on the topics in my published research. I encourage people with the above backgrounds to consider contributing their knowledge to speech recognition and speaker authentication research. I also encourage colleagues in the field of speech recognition and speaker authentication to extend their knowledge to the above fields in order to achieve breakthrough research results. This book does not include those fundamental topics which have been very well introduced in other textbooks. This author assumes the reader has a basic understanding of linear systems, signal processing, statistics, and pattern recognition.
This book can also be used as a reference book for government and company officers and researchers working in information technology, homeland security, law enforcement, and information security, as well as for researchers and developers in the areas of speaker recognition, speech recognition, pattern recognition, and audio and signal processing. It can also be used as a reference or textbook for senior undergraduate and graduate students in electrical engineering, computer science, biomedical engineering, and information managementMultivariate Statistical Analysis and One-Pass Vector Quantization.
Principal Feature Networks for Pattern Recognition.
Non-Stationary Pattern Recognition.
Robust Endpoint Detection.
Detection-Based Decoder.
Auditory-Based Time Frequency Transform.
Auditory-Based Feature Extraction and Robust Speaker Identification.
Fixed-Phrase Speaker Verification.
Robust Speaker Verification with Stochastic Matching.
Randomly Prompted Speaker Verification.
Objectives for Discriminative Training.
Fast Discriminative Training.
Verbal Information Verification.
Speaker Authentication System Design.