A RBF Neural Network Algorithm for Blind Source Separation of Linear Mixing Signals

This paper presents a radial basis function (RBF) neural network approach to blind source separation in linear mixture. After calculating center value vector and width value vector, weight value vector that is deduced by maximizing entropy (ME) of cost function is calculated in this RBF neural network. This cost function results in the independence of the outputs with desirable moments such that the original sources are separated properly. Simulation results show that the separation time is reduced and the separation effect is very good. Compared with ME of algorithm, the effect of this algorithm is better.