Equalization of satellite UMTS channels using RBF networks

The paper proposes a radial basis function (RBF) neural network (NN) approach to satellite universal mobile telecommunication system (S-UMTS) channel equalization. Two main problems arise in S-UMTS communications: (i) non-linear distortions which are caused by the onboard nonlinear power amplifiers, and (ii) multipath propagation. This paper shows that RBF networks are particularly well suited to overcome these problems. The paper presents several simulation results which show that RBF networks outperform classical equalization techniques for different configurations of the mobile channel (mobile speed, noise level, number of paths, etc.).