Statistical signal processing. Modelling and estimation. Transl. from the French by Janet Ormrod. Incl. 1 CD-ROM (Q1347436)

This book presents an introduction to statistical signal processing. It mainly deals with the modelling and spectral estimation of wide sense stationary processes, and their filtering. It also includes the techniques that generalize the methods only involving second-order statistics. The book progresses from the definition of random processes to adaptive estimation. In order to supply the reader with the desirable theoretical bases and to allow one to make progress in the discipline, most of the results presented in the book are carefully justified. There are sixteen chapters and Appendices A to Z in the book. Except Chapter 1, an introductory chapter, carefully chosen exercises are provided for each chapter. The CD-ROM supplied with the book contains MATLAB programs in HTML format and is intended to provide simulation examples (program listings and simulation results). In addition, it also contains probability course notes that are slightly too long to be presented as a mere appendix. In this book, Chapter 1 gives its motivation, presents a few classical problems in statistical signal processing, and describes its contents. In Chapter 2, after some basics about stochastic processes, special attention is paid to wide sense stationary stochastic processes. In Chapter 3, their power spectrum is defined, in Chapter 4 the notion of spectral representation is studied, and in Chapter 5 their filtering is examined. Chapter 6 covers some types of processes that are of particular interest for signal processing, including Gaussian processes, Poisson processes, cyclostationary processes and circular processes. In Chapter 7, some nonlinear or time-dependent transforms that are especially important in electronics and in signal communications are presented as examples of nonlinear transforms of a process. Chapter 8 is devoted to examining in particular under what conditions a process may be seen as the output of a causal filter with a white noise input as well as the problem of finite order linear prediction. Chapter 9 studies two particular filtering techniques: Wiener filtering and Kalman filtering based on the results of the theory of linear prediction. Chapter 10 investigates processes whose spectrum is given by a rational function. Chapter 11 considers the problem of the spectral identification of a process from knowledge of its first Fourier coefficients. Chapter 12 covers the non-parametric spectral estimation of stationary processes. In Chapter 13, it is shown how the parametric estimation of rational spectra using a minimum mean square approach may be undertaken simply. The case of Gaussian processes is considered. Chapter 14 presents the notion of a cumulant spectrum that generalizes the conventional definition of a spectrum as the Fourier transform of the autocovariance function. In Chapter 15, Bayesian estimation methods are covered and Monte Carlo techniques are used to solve integration and maximization problems that appear in Bayesian estimation. Chapter 16 is devoted to stochastic optimization algorithms for estimating the parameters associated with non-stationary phenomena. The emphasis is on least mean square algorithms and recursive least square algorithms. Some basics of probability and proofs of several theorems in the book are provided in the appendices. This book is intended for graduate students, especially for students both in telecommunications and applied statistics. It can also serve as an excellent reference book for engineers, researchers, and professors interested in statistical signal processing. I have found that the book is very helpful.