Adaptive channel equalization using counter propagation networks

Abstract

Effective channel equalization in a digital communication system is essential for error free data transmission over dispersive channels. Conventional adaptive equalization uses the linear LMS algorithm for this purpose. However, the performance of this equalizer degrades severely when the channel is nonlinear in nature; which is true in most of the practical situations. The equalizer performance can be improved by the use of nonlinear processing in an artificial neural network. We propose a counter propagation network (CPN) based adaptive channel equalizer for the task of channel equalization. It has been shown by extensive computer simulation that this equalizer performs much better than a LMS based equalizer. Over a wide range of SNR and eigenvalue ratio (EVR), the CPN based equalizer outperforms the LMS equalizer in terms of MSE for channels possessing linear as well as nonlinear characteristics.<>