Classifying Heart Rate Variability Data using Multitaper Spectrum Analysis

Sammanfattning: Heart rate variability (HRV) is the variation between two consecutive heartbeats. The irregular variability in this interval can indicate different health issues such as stress. The goal of this project is to correctly classify if a HRV signal comes from a resting state or a state which is affected by stress related stimuli. The analysis will be conducted using non-parametric multitaper spectrum analysis in the frequency bandwidth, 0.12-0.4 Hz. Two different multitaper methods will be tried; Welch method and Thomson method. For the binary classification of the HRV signal, it was assumed that there was a difference in energy distribution. In the pair-wise classification the assumption was instead that there was a difference in total energy. The highest and most trustworthy binary classification of the methods was 65% for Welch and 69% Thomson, with the assumption of there being difference in energy distribution. For the pair-wise classification, it was 74% and 77%, respectively.