Statistical approach to blind adaptive channel equalizations

Abstract

In the performance of digital communication systems, a fundamental limiting factor is the intersymbol interference(ISI) caused by bandlimiting or by frequency selective fading as, for example, in telephone or digital microwave radio channel. Removal of ISI is usually accomplished by use of adaptive equalizer at receiver. Conventionally, the adaptive equalization is initiated by sending a training sequence which helps the equalizer adapt to the unknown channel characteristics. Unlike the conventional adaptive equalization schemes, the blind equalization algorithms solely draw upon the received sequence and a priori knowledge of the statistics of the transmitted sequence. This dissertation presents statistical approaches to the blind equalizations in digital communication channels. There has been an increasing interest in applying the higher-order (higher than two) statistics to the blind equalization because the higher-order statistics not only reveal amplitude information about a process, but also reveal phase information while the second-order statistics (i.e., correlations) are phase blind. Basically, these approaches are accounted for the minimization of the cost functions defined by the necessary (higher-order) statistics of the transmitted, received, or equalizer output sequence for the blind equalizations. The problems considered in this dissertation are the designs of the cost function for the blind equalization, their optimization via the stochastic gradient method, and convergence analysis of the proposed algorithms. The contributions of this dissertation are (i) Convergence analysis of FIR (finite impulse response) CMA (constant modulus algorithm) blind equalizers, (ii)Initialization scheme of FIR CMA blind equalizers,(iii)Blind equalization via the higher-order statistics, and (iv) Blind equalization via absolute mean. In Chapters 2 and 3, the convergence properties and initialization of FIR CMA equalizers are analyzed. In Chapter 4, new blind iden...