High-Resolution Time-Frequency Energy Features of HRV signals using the SPWVD and the STFT-Spectrogram

Information within heart rate variability (HRV) signal allows detecting severe cardiovascular pathologies such as acute myocardial infarction, congestive heart failure, etc. The frequency-domain content of HRV signals is quantified as Low-Frequency (LF) and High-Frequency (HF) spectral components in relation to sympathetic and parasympathetic autonomic nervous system activities, respectively. Time-frequency analysis can effectively supersede spectral analysis to reveal characterizing non-stationary features within HRV signals over LF and HF bandwidths. Therefore, in this paper, we defined new time-frequency (TF) energy features that can discriminate between various cardiovascular pathological cases by analyzing their respective HRV signals. Indeed, we defined TF Normalized Energy ratios calculated over both LF and HF bandwidths, denoted as NELF and NEHF, within TF representations calculated by the Smoothed Pseudo Wigner-Ville distribution (SPWVD) and the spectrogram (SP) for HRV signals generated from ECG records (100 up to 109) of the MIT-BIH Arrhythmia Database. As a performance evaluation, We found out that the SPWVD yields better localization of non-stationary components of the analyzed HRV signals within the TF plane in comparison to the SP.