Time-frequency analysis of heart rate variability during stress testing using “a priori” information of respiratory frequency

A parametric decomposition of the smoothed Wigner-Ville distribution is applied to estimate the instantaneous frequency and amplitude of the LF and HF components of HRV during stress testing. It is assumed that the instantaneous frequencies of the LF and HF components may vary linearly with time during stress testing, and that the frequency of the HF component can be approximated by the respiratory frequency. The effect of the inclusion of "a priori" information of respiratory frequency on the estimation of HRV parameters is studied. Results on a simulation study show that the inclusion of "a priori" information does not affect significantly the estimation of the LF parameters. However, the SD of the estimation error of the HF amplitude is reduced at the expense of introducing a bias for high SNRs while both the mean and SD of the HF amplitude error are decreased for lower SNRs (2.6%plusmn1.9% against 3.1%plusmn2.6%for a SNR of 20 dB and 7.6%plusmn5.8% against 12.0%plusmn9.7%for a SNR of 10 dB).