Fuzzy neural network approach for 2-D direction finding in multipath environments

Abstract

A fuzzy neural network (FNN) approach for estimating the two-dimensional (2-D) direction of a radiating source in coherent multipath environments via a 2-D passive sensor array is presented. Outputs of the array are preprocessed by a judiciously constructed reference-point preprocessing scheme to produce proper inputs for an FNN. The FNN then uses these preprocessed outputs to produce the estimate of the source direction. Since the preprocessed outputs preserve enough information about the source direction and the FNN maps the preprocessed outputs into the source direction with high accuracy, the FNN approach exhibits good estimation performance. By computer simulations, it is found that the FNN approach drastically outperforms the spatial smoothing 2-D MUSIC algorithm.