Linear estimation and detection in Krylov subspaces (Q2642226): Difference between revisions

The book is focused on the foundations of linear estimation theory which is essential for effective signal processing. One major area in this theory is reduced-rank estimation where optimal linear estimators are approximated in low-dimensional subspaces, e.g., in order to reduce the noise in overmodeled problems, enhance the performance in case of estimated statistics, and/or save computational complexity in the design of the estimator which requires the solution of linear equation systems. The presented book provides a comprehensive overview over reduced-rank filters where the main emphasis is put on matrix-valued filters whose design requires the solution of linear systems with multiple right-hand sides. In particular, the multistage matrix Wiener filter, i.e., a reduced-rank Wiener filter based on the multistage decomposition, is derived in its most general form. In the first part of the book, a comprehensive overview of several key methods like reduced-rank signal processing and Krylov subspace methods of numerical mathematics, are given. Based on the derivation of the multistage Wiener filter in its most general form, the relationship between statistical signal processing and numerical mathematics is presented. In numerical mathematics, iterative block Krylov methods are very popular techniques for solving systems of linear equations with multiple right-hand sides, especially if the systems are large and sparse. Besides presenting a detailed overview of the most important block Krylov methods in Chapter 3, which may also serve as an introduction to the topic, their connection to the multistage matrix Wiener filter is revealed in this book. Especially, the reader will learn the restrictions of the multistage matrix Wiener filter which are necessary in order to end up in a block Krylov method. This relationship is of great theoretical importance because it connects two different fields of mathematics, statistical signal processing and numerical linear algebra. Besides the book mainly addresses readers who are interested in the theory of reduced-rank signal processing and block Krylov methods, it includes also practical issues like efficient algorithms for direct implementation or the exact computational complexity in terms of the required number of floating point operations. If the reader is not interested in these practical aspects, Sections 2.2, 4.3, and 4.4 of this book can be skipped. In the second part, the theory is applied to iterative multiuser detection receivers (Turbo equalization) which are typically desired in wireless communication systems. The investigations include exact computational complexity considerations and performance analysis based on extrinsic information transfer charts as well as Monte-Carlo simulations. The book covers the application of the proposed linear estimators to a detection problem occurring at the receiver of a digital communication system. An iterative (Turbo) multiuser detector is considered where users are separated via spread spectrum techniques. Besides using Monte Carlo simulations, the communication system is investigated in terms of the expected iterative estimation error based on extrinsic information transfer charts. It should be mentioned that the extrinsic information transfer characteristics that are shown in these charts, are calculated in a semianalytical way as derived in Section 6.1.2. The content of the book will be of interest for all students, specialists and researchers, working in the area of circuit theory, signal processing and communications engineering, especially on the multistage Wiener filtering and Krylov methods.