Theoretical insights into the Phase Rectified Averaging method and application to HRV analysis

Heart rate variability (HRV) signals are processed using the new time-frequency representation (PRSA-TFR) defined based on the Phase-Rectified Signal Averaging (PRSA) method. PRSA is a technique which enhances quasi-periodic components in nonstationary signals and thus improves frequency estimation. The PRSA-TFR is obtained by applying the PRSA method to sliding windows along univariate signals. Our aim is to characterize the deviation of PRSA-TFR of HRV singal during supine and tilt position from that of white Gaussian noise. This deviation can be used as a new tool to quantify the changes in sympathovagal balance with out needing to predetermine fixed spectral boundaries. First, we derive the probability density function of the energy distributed in the PRSA-TFR for a white Gaussian noise. Then, the Battacharya distance is used to evaluate the deviation of HRV PRSA-TFR from that of a Gaussian nosie. The HRV PRSA-TFR deviation is assessed separately for supine and tilt positions. Synthetic and real HRV signals of short-term recordings are analyzed based on this new tool. The obtained results are compared with those obtained with a classical spectrao method.