Object recognition in ground surveillance doppler radar by using bispectrum-based time-frequency distributions

Performance estimation and comparative study of different kinds of time-frequency distributions (TFDs), especially Wigner-Ville distribution (WVD), as well as bispectrum-based distributions: Wigner-bispecrum (WBD), parametric (PBD) and non-parametric (NPBD) obtained by real non-stationary echosignal measuring in ground surveillance Doppler radar are considered for solving moving object recognition and classification problems. Automatic target recognition (ATR) performance has been examined by using information features contained both in the TFDs and cepstral coefficients. Experimental results performed by ground surveillance radar operating in millimeter wavelength for walking human, group of walking humans, bicyclist, vehicles and vegetation clutter show that bispectrum-based approach seems to be attractive tool for ATR systems comparing to the conventional energy-based approaches.