Reducing Bayes equalizer complexity: a new approach for clusters determination

Abstract

A new strategy for channel equalization in digital communication is presented. In this approach, the clustering problem is treated analytically. We propose a systematic Bayesian classification using a Gaussian approximation of the probability density function for each cluster. The quality of the approximation depends on the number of clusters considered. We show analytically that we can obtain the Bayes equalizer performance if we use the maximum number of clusters and we can obtain the Wiener equalizer performance if we use only two clusters (binary signal case). Some computational simulations illustrate the power of the presented strategy.