Speech signal processing using a time-frequency distribution of Cohen's class

It is well known that the standard methods for spectral estimation of speech signals, such as spectrograms, suffer from trade-offs between time and frequency resolution. These methods usually have difficulty isolating closely spaced formants and tracking rapidly changing spectral peaks and fundamental frequency (pitch). The authors use a time-frequency distribution with compound kernel function of Cohen's class to visualize these temporal features in both the time and frequency domains simultaneously. It shows the time-varying characteristics of the speech signal more precisely.<
>