Channel equalization by feedforward neural networks

Abstract

A signal suffers from nonlinear, linear, and additive distortion when transmitted through a channel. Linear equalizers are commonly used in receivers to compensate for linear channel distortion. As an alternative, nonlinear equalizers have the potential to compensate for all three sources of channel distortion. Previous authors have shown that nonlinear feedforward equalizers based on either multilayer perceptron (MLP) or radial basis function (RBF) neural networks can outperform linear equalizers. In this paper, we compare the performance of MLP vs. RBF equalizers in terms of symbol error rate vs. SNR. We design a reduced complexity neural network equalizer by cascading an MLP and a RBF network. In simulation, the new MLP-RBF equalizer outperforms MLP equalizers and RBF equalizers.