Estimating radar clutter distributions via neural networks

Abstract

A new approach to the solution of constant false alarm rate (CFAR) problems in complex, unknown environment is proposed. An adaptive system is described in which a neural network properly transforms the input random process which describes the clutter amplitude. The network proposed is a modified neural network (Tchebychev neural network), specifically designed for the problem at hand, whose neurons implement Tchebychev polynomials up to a proper order. Experimental results are presented referring to Weibull and Rayleigh input distributions.