Adaptive Time-Frequency Interbeat

In this paper, we address the nonstationary problem to analyze the cardiovascular control system based on beat- to-beat heart rate variability (HRV) time series. A novelty information here is to perform HRV time-frequency analysis without "a priori" knowledge of nonstationarity. This is achieved by adaptive interactions that will decompose the HRV signal in sub-signals with well-defined time-frequency parameters. These may be used to reconstruct the signal in sub-bands or might be used in screening tests. To validate our method, we carried out the experiments with both artificial and real HRV intervals. As an application test, we consider a simple case of screening, where experiments with a database consisting of beat-to-beat samples derived from 20 normal and 15 cardiac heart failures (CHF) yielded an overall classification accuracy of 98.33 %.