Efficient Architecture for Bayesian Equalization Using Fuzzy Filters

Abstract

A normalized Bayesian solution is derived for digital communication channel equalization which uses estimates of scalar channel states. This equalizer is termed as a normalized Bayesian equalizer with scalar channel states (NBEST). The relationship between the NBEST and fuzzy equalizers is derived and computational aspects of fuzzy equalizers are investigated using different types of fuzzy basis functions. It is shown that the fuzzy equalizer in general demands much lower computational complexity than the optimum equalizer. Ways to further reduce the computation complexity of fuzzy equalizers is proposed and their performance evaluated. A novel scheme to select a subset of channel states close to the received vector, resulting in considerable reduction in the computational complexity, is also proposed. A fuzzy equalizer with this modified membership function is shown to perform close to the Bayesian equalizer.