Copula-based positive time-frequency distributions of the phonocardiogram

This paper introduces a method for time-frequency analysis of the phonocardiogram. The proposed method is a modified version of the copula-based technique, which constructs positive time-frequency distributions, so as to be suitable for the analysis of periodically stationary random processes. The copula-based time-frequency distributions have some desirable properties: positivity, correct marginals and low computational cost. We use averaged spectrogram as a template in copula-based technique in order to construct a positive time-frequency distribution of the phonocardiogram. Results clearly show that the copula-based time-frequency distributions have much better resolution in time and frequency domain than the spectrogram does. Positivity and high resolution depend on the marginals of the distributions will offer excellent potential for estimating local parameters of the distributions, like a instantaneous frequency or instantaneous equivalent bandwidths using Renyi's divergence formulated by the authors.