A new time-frequency distribution and its application to speech signal processing

Abstract

A kernel function for the time-frequency distribution of Cohen's (1989) class, which is the product of the Choi-Williams (1989) and the Margenau-Hill (1961) kernels, is proposed. Specific types of signal (sinusoidal signals, chirp signals and others) are analyzed using the distribution and compared with results for the Wigner-Ville and the Choi-Williams distributions. It is found that the present distribution does not indicate spurious intensity in the regions where the other distributions do. The three distributions are compared numerically for the above signals and for speech signals, in order to show the advantages of the present distribution.<>