Correlation Between Heart Rate Variability Features

Several features for heart rate variability (HRV) are being proposed in many studies. However, it is unclear whether new methods are contributing to knowledge or just duplicating the information represented by existing methods. Therefore, in this study, the correlation between different traditional and modern approaches to HRV analysis has been analyzed. The HRV features were extracted from signals of four classes of subjects acquired from PhysioNet databases. Three correlation coefficients, namely Pearson's correlation (PearCorr), Kendall's rank correlation (KendCorr), and distance correlation (DistCorr), have been used. PearCorr is sensitive to only linear relationships, whereas KendCorr and DistCorr are sensitive to both linear and nonlinear associations between variables. The experimental results suggest that traditional time-domain based features show a significant correlation with Poincaré plot-based features. The frequency-domain features show a considerable relationship with entropy-based complexity measures of HRV. The modern approaches to complexity analysis, such as heart rate asymmetry indices and phase entropy, do not show a significant correlation with other existing HRV features.