Time-frequency analysis of frequency-coded signals

Abstract

Application of time-frequency based techniques on the analysis of frequency-coded signals is considered. Their performance are compared on the frequency-coded signals with optimal Costas (1984) sequence. For signals with constant frequencies, within one coding interval (CW signals), satisfactory results can be achieved by using the spectrogram with appropriate lag window. For time-varying (including linear frequency modulated) elementary signals the spectrogram exhibits poor time-frequency resolution and estimation accuracy. The estimation accuracy and time-frequency concentration improvement for this kind of signals can be achieved by using the cross-terms reduced version of the Wigner distribution, the S-method. In the case of linear frequency modulated elementary signals very accurate results can be obtained by using adaptive fractional Fourier transform, introduced in this paper. It is based on the concentration measure of the time-frequency representations. In order to improve the estimation accuracy in a high noise environment a post median filtering is applied to the resulting, estimate.