Improved interpolation of unevenly sampled heart rate signals

Abstract

In the heart rate variability (HRV) analysis, based on power spectral density (PSD), there are two major problems: The accessibility of information (beat timing) and the uneven sampling of the information carrying signal. The first problem has been addressed through the heart rate (HR), the heart period (HP), and recently we have proposed the heart timing (HT) signal. Interpolation has been done with linear or cubic spline methods. In this work, we propose an alternative, that is a combination of the adaptive weights preconditioning method and the method of conjugate gradient for the solution of positive definite linear systems. We show how this reconstruction algorithm allows the recovery of a sequence of regular spaced samples and its spectrum from the proposed HT, with higher precision (lower low-pass effect) than other methods.