Algorithms for Statistical Signal Processing

The field of digital signal processing (DSP) has expanded rapidly over the past three decades. During the late sixties and seventies, we witnessed the development of the basic theory for digital filter design and the development of computationally efficient algorithms for evaluating the Fourier transform, convolution, and correlation. During the past two decades, we experienced an explosion in DSP applications spurred by significant advances in digital computer technology and integrated-circuit fabrication. In this period, the basic DSP theory has expanded to include parametric signal modeling, with applications to power spectrum estimation and system modeling, adaptive signal processing algorithms, multirate and multidimensional signal processing, and higher-order statistical methods for signal processing. With the expansion of basic DSP theory and the rapid growth in applications (spurred by the development of fast and inexpensive digital signal processors), there is a growing interest in advanced courses in DSP covering a variety of topics. This book was written with the goal of satisfying, in part, the resulting need for textbooks covering these advanced topics. Most of the material contained in this book was first published in 1992 by the Macmillan Publishing Company, in a book entitled Advanced Digital Signal Processing (which went out of print in 1997). This new book differs from the earlier publication by the inclusion of a new chapter (Chapter 7) on QRD-based fast adaptive filter algorithms, and the deletion of a chapter on multirate signal processing. The other chapters have remained essentially the same. The major focus of this book is on algorithms for statistical signal processing. Chapter 2 treats computationally efficient algorithms for convolution and for the computation of the discrete Fourier transform. Chapter 3 treats linear prediction and optimum Wiener filters; included in this chapter is a description of the Levinson-Durbin and Schur algorithms. Chapter 4 considers the filter design problem based on the least-squares method and describes several methods for solving least squares problems, including the Givens transformation, the Householder transformation, and singular-value decomposition. Chapter 5 treats single-channel adaptive filters based on the LMS algorithm and on recursive least-squares algorithms. Chapter 6 describes computationally efficient recursive least-squares algorithms for multichannel signals. Chapter 7 is focused on the uses of signal flow graphs for deriving computationally efficient adaptive filter algorithms based on the QR decomposition. Chapter 8 deals with power spectrum estimation, including both parametric and nonparametric methods. Chapter 9 describes the use of higher-order statistical methods for signal modeling and system identification. Although the material in this book was written by six different authors, we have tried very hard to maintain common notation throughout the book. We believe we have succeeded in developing a coherent treatment of the major topics outlined in the preceding overview. Chapter 1 provides an introduction to selected basic DSP material that is typically found in a first-level DSP text, and also serves to establish some of the notation used throughout the book. In our treatment of the various topics covered herein, we generally assume that the reader has had a prior course on the fundamentals of digital signal processing. The fundamental topics assumed as background include the z-transform, the analysis and characterization of discrete-time systems, the Fourier transform, the discrete Fourier transform (DFT), and the design of FIR and IIR digital filters. John G. Proakis Charles M. Rader Fuyun Ling Chrysostomos L. Nikias Marc Moonen Ian K ProudlerProakis, John G. is the author of 'Algorithms for Statistical Signal Processing' with ISBN 9780130622198 and ISBN 0130622192.