Modeling and estimation of time-varying heart rate variability during stress test by parametric and non parametric analysis

A methodological framework for simulating real-like HRV during stress test with controlled spectral properties has been developed with the purpose to assess SPWVD and time-variant AR analysis. For each method results have been evaluated computing the estimation error for LF and HF components during all the test (mean error of the order of 10% in all case) and a direct comparison based on the correlation between estimate and original ACF yields very high values (rho_WV=0, 99 and rho_AR=0, 94). In real data analysis bothmethods highlightan inversion of the relative spectral balance around the apex of exercise which passes from LF to HF prevalence.