Adaptive SPWV distribution with adjustable volume 2-D separable kernel

Abstract

In practical time-frequency (T-F) analysis of nonstationary signals associated with the use of the smooth pseudo Wigner-Ville distribution (SPWV), an important issue is a choice of the smoothing window. To address this problem, a fast sub-optimal technique is proposed, relating the window length to the approximate slope of the frequency modulation. The results obtained by using this automated procedure are comparable to results given by supervised decompositions and are more (locally) adapted. The developed method has been used in processing of radar returns. In order to objectively assess the performance of the transforms, several measures can be used. For signals with unknown modulation laws, cost functions are available, including measures based on entropy function. Their application to evaluation of the obtained T-F distributions is demonstrated.