Functional Link Neural Network Based NARMA Channel Modelling

In a typical mobile environment because of the high speed of the terminals the channel coherence time becomes smaller and the assumption that the channel is constant within a frame no longer holds. To aid the channel estimation for coherent detection in such a situation, data aided estimation becomes inefficient because of the requirement of high density of reference symbols in a frame. Previously prediction of mobile channel based on ARMA model is established with the cost of high computational requirement to solve the Yule-Walkar equations. We investigated the performance of some Artificial Neural Networks (ANN) to model Rayleigh fading channel as ARMA, NARMA and AR process. Modified Functional Link Neural Network (FLNN) with nonlinear functional expansion is found to be suitable for modelling the Jack´s tap gain process.