Electronic Thesis and Dissertation Repository


Master of Science


Computer Science


Lizotte Daniel


Recently, among various analysis methods of physiological signals, automatic analysis of Electrocardiogram (ECG) signals, especially heart rate variability (HRV) has received significant attention in the field of machine learning. Heart rate variability is an important indicator of health prediction and it is applicable to various fields of scientific research. Heart rate variability is based on measuring the differences in time between consecutive heartbeats (also known as RR interval), and the most common measuring techniques are divided into the time domain and frequency domain. In this research study, a classifier based on analysis of HRV signal is developed to classify different activities including sleep, exam, and exercise. The performance of the classifier is improved using a novel feature construction approach named as baseline assisted classifier.

ECG data are collected from 39 subjects and RR intervals are derived from ECG data using Firstbeat analysis software to compute HRV metrics. These metrics are utilized as features in a logistic regression, SVM, decision tree, random forest classifiers. Performance of all classifiers is assessed by leave one person out cross-validation technique. Features are derived by statistical time domain method from HRV segmentation during 5-minutes recording. Using a combination of 5-minutes segmentation feature vector and 5-minutes segmentation feature vector of sleep record results in a median area under the receiver operating curve (AUC) of 88% for sleep and 74% for the exam on leave one person out cross-validation test set data by SVM classifier. These

results demonstrate that adding a baseline feature vector of sleep data improves the classification accuracy and classification AUC accuracy of almost all classifiers from HRV measures, and tracking of activity can be achieved by measuring the HRV signal.