Algorithms for Communications Systems and their Applications, 2nd Edition

Chapter 15Self‐training equalization

By the expression self‐training equalization we refer to channel equalization techniques by which we obtain the initial convergence of the parameters of an adaptive equalizer without resorting to the transmission of a training sequence. Although these techniques generally achieve suboptimum performance at convergence, they are applied in many cases of practical interest, where the transmission of known training sequences is not viable, as for example in broadcast systems, or may result in an undesirable increase of system complexity [1–6]. In these cases, it is necessary to consider self‐training of the adaptive equalizer, where the terms for adjusting the coefficient parameters are obtained by processing the received signal. Usually, the receiver performs self‐training of the equalizer by referring to the knowledge of the input signal characteristics, for example the probability distribution of the input symbols.

The subject of self‐training equalization in communications systems has received considerable attention since the publication of the article by Sato [7] in 1975; the proposed algorithms have been alternatively called self‐recovering, self‐adaptive, or blind.

15.1 Problem definition and fundamentals

With reference to the discrete‐time equivalent scheme of Figure 7.15 shown in Figure 15.1, we consider a real‐valued discrete‐time system given by the cascade of a linear channel , in general with non‐minimum phase transfer function ...