On cross-components in the Wigner distribution

Abstract

The origin and geometrical characteristics of auto- and cross-components in the Wigner distribution (WD) are investigated utilizing a probe signal. Techniques for eliminating cross-components are reviewed and compared. These methods include local smoothing of the WD using Gaussian and other kinds of smoothing windows, adaptive variation of the smoothing window parameters according to the signal's structure, and the use of singular value decomposition. All the techniques discussed involve a tradeoff between the cross-components and the time-frequency resolution. Ways of avoiding this tradeoff are discussed