Analysis of the Impact of Inter-Beat-Interval Interpolation on real-time HRV Feature Estimation for e-Health Applications
Résumé
Heart rate variability (HRV) has proven to be one of the most promising indicatorof many physiological and psychological states. Thanks to great innovations in wearabledevices, HRV is now measurable by simple sensors remotely connected via wirelessnetworks to computers or smartphones. However, these sensors aren’t as precise asthe gold standard Electrocardiographs (ECG) used in hospitals. Errors during thetransmission or acquisition may deteriorate signal’s quality and considerably affect HRVfeatures. These errors are not acceptable for a precise HRV analysis potentially used fordiagnosis. Therefore, in this study, we use four different interpolation methods (NearestNeighbour - NN, Linear, Shape-preserving piecewise cubic Hermite - Pchip and cubicspline) that help tackle the problem of missing RR values. We then investigate theireffect on HRV analysis in order to quantify the estimation error allowing to choosethe best interpolation method. The main particularity of this study is the real-timeapproach to data interpolation and HRV analysis. We observed that some interpolationmethods behave differently as missing values’ percentage grows. Some being moresuitable for RR timeseries with a greater number of missing data. The study alsosuggests that interpolation may have a greater impact on some HRV features comparedto others. Finally, in order to achieve maximum performance, we propose to adaptinterpolation method to both missing values’ percentage and targeted HRV feature.
Origine : Fichiers produits par l'(les) auteur(s)