Ground moving target classification by using DCT coefficients extracted from micro-Doppler radar signatures and artificial neuron network

Abstract

A novel approach to ground moving targets classification by using information features contained in micro-Doppler radar signatures is presented. Suggested approach is based on using discrete cosine transform (DCT) coefficients extracted from radar signature as a classification feature and multilayer perceptron (MLP) as a classifier. Proposed pattern classification algorithm was tested by utilizing experimental data measurements performed by ground surveillance Doppler radar system for four radar target classes as single moving human, groups of two and three moving persons and vegetation clutter. Suggested approach provides the probability of classification equal to 86%