Moving target classification in ground surveillance radar ATR system by using novel bicepstral-based information features

A novel bicepstrum-based strategy is suggested for moving target recognition and classification by ground surveillance Doppler radars. Bicepstral coefficients extracted from non-stationary backscattered radar signals are used as the information features in automatic target recognition (ATR) system for solving a problem of moving human recognition and classification. ATR performance is studied by using Gaussian mixture model (GMM) and maximum likelihood (ML) rule. Experimental results obtained by ground surveillance homodyne radar operating in millimeter range wavelengths and continuous mode are represented and discussed. Classifier performance is examined in real-life conditions for walking human and group of walking humans in vegetation clutter environment.