HRV spectral estimation based on constrained Gaussian modeling in the nonstationary case

Abstract

The heart rate variability (HRV) spectral parameters are classically used for studying the autonomic nervous system, as they allow the evaluation of the balance between the sympathetic and parasympathetic influences on heart rhythm. However, this evaluation is usually based on fixed frequency regions, which does not allow possible variation, or is based on an adaptive individual time dependent spectral boundaries (ITSB) method sensitive to noisy environments. In order to overcome these difficulties, we propose the constrained Gaussian modeling (CGM) method that dynamically models the power spectrum as a two Gaussian shapes mixture. It appeared that this procedure was able to accurately follow the exact parameters in the case of simulated data, in comparison with a parameter estimation obtained with a rigid frequency cutting approach or with the ITSB algorithm. Real data results obtained on a classical stand-test and on the Fantasia database are also presented and discussed.