UMTS channel equalization using neural networks: natural gradient algorithm

Neural networks (NN) have been widely used in adaptive signal processing last decade. Their adaptive properties and nonlinear characteristics make them powerful tools to overcome difficult nonlinear problems. Satellite communications channels have nonlinear complex gain which makes the decision boundaries between different symbols nonlinear. NNs have been successfully used for satellite channel equalization. In this paper we present a multilayer neural network equalizer applied to UMTS channel equalization. We propose to use the natural gradient algorithm instead of the ordinary gradient in order to enhance the convergence properties of the NN.