Selection of a time-frequency representation for analysis of the first heart sound signal

A simulated first heart sound (S1) signal was produced based on the hypothesis that S1 is generated not only by valvular vibration but also by myocardial contraction. This simulated S1 was then used to optimize five time-frequency representations (TFR): the spectrogram, the time-varying autoregressive (AR) modeling, the Bessel distribution (BD), the cone-kernel distribution (CKD), and the binomial transform reduced interference distribution (BTRID). The results of comparing each of the five representations with the theoretical distribution of the simulated S1 show that each TFR has its own limitations. However, the CKD is the best technique for this specific application.