Algoritmo LMS com Passo Adaptativo para Utilização em Equalizadores Cegos Bayesianos<br />10.14209/jcis.2001.8

Abstract

A blind equalizer that performs joint estimation of channel and symbols is proposed. Its origin comes from two blind equalization schemes proposed in [1] which are based on the Kalman filter and the LMS algorithm. The performance of the most complex equalizer under multipath channel models is much higher than the least complex one. A better trade-off between performance and computational complexity was searched through an investigation of strategies to improve the LMS performance. An original result obtained from this investigation is the new blind equalization algorithm here proposed which uses LMS filters with adaptive step-size parameter based on a metric used in another work for blind detection of equalization errors. An extensive performance evaluation of these equalizers under fast frequency-selective fading channels is presented. The results reached by the proposed algorithm show a remarkable improvement in the performance compared with the scheme based on standard LMS, and a reduction of the complexity compared with the one based on Kalman filters.