Improved Chirplet time-frequency atom for signal decomposition and its application to micro-Doppler analysis

Radar micro-Doppler (m-D) is the phenomenon of the observed micro-motions on top of the bulk main Doppler component of a target's motion. The micro-Doppler features of radar target returns were extracted through various time-frequency (T-F) analyses. The new method for T-F analysis is based on dictionary decomposition, in which the time-frequency signal is decomposed into a combination of time-frequency atoms. In view of the weak decomposition performance of Chirplet time-frequency atoms for sinusoidal type nonlinear frequency modulated (NLFM) signals, an Improved Chirplet (IChirplet) T-F atom is proposed. Firstly, the mismatch problem of T-F distribution of Chirplet time frequency atom to sinusoidal type is analyzed theoretically. Secondly, sinusoidal frequency modulation factor is added to Chirplet time frequency atom to make the time frequency curve of the atom produce the flexural performance of the sine like shape. Finally, the genetic algorithm is used instead of matching pursuit algorithm to improve the efficiency of the atomic search. The simulation results show that the decomposition performance of the IChirplet atom for the NLFM signal has greatly improved compared with the Gabor, Chirplet, and FM
^m
let atoms. We apply the IChirplet T-F atom to m-D signals analysis. The simulation results show that it is capable of separating the micro-motions from translation in IChirplet parameter domain.