Time-frequency analysis of the non-stationary multi-frequency signals by using parametrical short-time bispectral density estimation

A new approach to time -frequency analysis of multi-component frequency modulated signals is represented. The suggested approach is based on the extraction of phase -coupled instantaneous frequency pairs containing in the broad-band Doppler signal spectrum. For this purpose, the computation of the sequence of short-time parametric bispectral density estimates of transient signal segments separated by a sliding window is used. The parametric autoregressive (AR) modeling has been employed to obtain high frequency resolution. Time-frequency distributions of the phase-coupled instantaneous frequencies are represented for non -stationary multi-component signal models as well as for experimental Doppler radar records obtained for signals backscattered by a moving human. The latter radar signatures demonstrate evolutionary behaviour of phase-coupled instantaneous frequency pairs in the time -frequency domain. The given results can be useful as new information features for automatic target radar recognition.