Unsupervised Adaptive Filtering, Blind Deconvolution / Edition 1

by Simon Haykin

Overview

A complete, one-stop reference on the state of the art of unsupervised adaptive filtering

While unsupervised adaptive filtering has its roots in the 1960s, more recent advances in signal processing, information theory, imaging, and remote sensing have made this a hot area for research in several diverse fields. This book brings together cutting-edge information previously available only in disparate papers and articles, presenting a thorough and integrated treatment of the two major classes of algorithms used in the field, namely, blind signal separation and blind channel equalization algorithms.

Divided into two volumes for ease of presentation, this important work shows how these algorithms, although developed independently, are closely related foundations of unsupervised adaptive filtering. Through contributions by the foremost experts on the subject, the book provides an up-to-date account of research findings, explains the underlying theory, and discusses potential applications in diverse fields. More than 100 illustrations as well as case studies, appendices, and references further enhance this excellent resource. Following coverage begun in Volume I: Blind Source Separation, this volume discusses:
* The core of FSE-CMA behavior theory
* Relationships between blind deconvolution and blind source separation
* Blind separation of independent sources based on multiuser kurtosis optimization criteria

Product Details

ISBN-13:	9780471379416
Publisher:	Wiley
Publication date:	04/06/2000
Series:	Adaptive and Cognitive Dynamic Systems: Signal Processing, Learning, Communications and Control , #24
Edition description:	Volume 2
Pages:	200
Product dimensions:	6.32(w) x 9.53(h) x 0.76(d)

About the Author

SIMON HAYKIN, PhD, is University Professor and Director of the Adaptive Systems Laboratory at McMaster University.

Table of Contents

Contributors vii

Preface xi

1 Introduction 1
Simon Haykin

1.1 Why Adaptive Filtering?  1

1.2 Supervised and Unsupervised Forms of Adaptive Filtering 2

1.3 Two Important Unsupervised Signal-Processing Tasks 3

1.4 Three Fundamental Approaches to Unsupervised Adaptive Filtering 6

1.5 Organization of Volume II 10

References 11

2 The Core of FSE-CMA Behavior Theory 13
C. R. Johnson, Jr., P. Schniter, I. Fijalkow, L. Tong, J. D. Behm, M. G. Larimore, D. R. Brown, R. A. Casas, T. J. Endres, S. Lambotharan, A. Touzni, H. H. Zeng, M. Green, and J. R. Treichler

2.1 Introduction 14

2.2 MMSE Equalization and LMS 22

2.3 The CM Criterion and CMA 41

2.4 CMA-Adapted-Equalizer Design Issues with Illustrative Examples 75

2.5 Case Studies 89

2.6 Conclusions 106

References 108

3 Relationships between Blind Deconvolution and Blind Source Separation 113
Scott C. Douglas and Simon Haykin

3.1 Introduction 113

3.2 Problem Descriptions 117

3.3 Algorithmic Relationships 122

3.4 Structural Relationships 129

3.5 Extensions 140

3.6 Conclusions 142

References 142

4 Blind Separation of Independent Sources Based on Multiuser Kurtosis Optimization Criteria 147
Constantinos B. Papadias

4.1 Introduction 148

4.2 Problem Formulation and Assumptions 150

4.3 Review: The Single-User Equalization Problem 154

4.4 Necessary and Su½cient Conditions for BSS 160

4.5 Unconstrained Criteria: The MU-CM Approach 162

4.6 Constrained Criteria: The MUK Approach 165

4.7 Numerical Examples 171

4.8 Conclusions 175

References 176

Index 181

Preface

UNSUPERVISED ADAPTIVE FILTERING

Volume II: Blind Deconvolution

PREFACE

In 1994 I edited a book on ``blind deconvolution,'' which presented an account of the various algorithms that had been developed essentially for solving the blind channel-equalization problem. The material presented in that book spanned a period of over 25 years, going back to the pioneering work of Robert Lucky in 1966 on the decision-directed mode of operating the least-mean-square algorithm and that of Y. Sato in 1975 on a blind channel-equalization algorithm that bears his name. These two pioneering contributions were followed by another pioneering contribution to blind channel equalization, namely, the constant-modulus algorithm that was developed independently by Godard in 1980 and Treichler and Agee in 1983. Subsequently, it was recognized that these three blind equalization algorithms are members of the family of Bussgang algorithms.

In 1994 Pierre Comon published a paper in a signal-processing journal on ``independent component analysis,'' which was followed by Tony Bell and Terry Sejnowski's 1995 paper in a neural computation journal on the Infomax (or, more precisely, the maximum-entropy) algorithm for blind signal separation. Although, indeed, work on the blind signal- separation problem could be traced to a much earlier paper by J. Herault, C. Jutten, and B. Ans that was published in 1985, it would be fair to say that Pierre Comon's paper and that of Tony Bell and Terry Sejnowski served as catalysts for raising the profile of research interests in blind source separation to the extent that the subject has become a ``hot'' area with potential applications in a variety of diverse fields.

Despite the fact that blind channel equalization and blind source separation have originated in their own somewhat independent ways, they are in actual fact intimately related to each other. Indeed, they constitute the two pillars of unsupervised adaptive filtering. By bringing them together under the umbrella of this new book, organized in two volumes, not only have we provided an up-to-date treatment of blind signal-separation and blind channel-equalization algorithms and their underlying theoretical formalisms but also opened an avenue for the cross-fertilization of new ideas. Volume I of the book covers blind source-separation algorithms, and Volume II covers blind deconvolution (i.e., blind equalization) and its relationship to blind source separation.

I would like to take this opportunity to express my deep gratitude to each and every one of my coauthors for making the writing of this unique two-volume work a reality.

Simon Haykin

Ancaster, Ontario, Canada

March 2000

Editorial Reviews

Covers blind equalization and its relationship to blind source separation. Provides an account of research findings, explains underlying theory, and discusses potential applications in diverse fields, with sections on the core of FSE-CMA behavior theory, relationships between blind deconvolution and blind source separation, and blind separation of independent sources based on multiuser kurtosis optimization criteria. The editor is affiliated with McMaster University. Annotation c. Book News, Inc., Portland, OR (booknews.com)

Booknews