Heart rate variability analysis using neural network models for automatic detection of lifestyle activities

• Published in: Biomedical signal processing and control Vol. 42; pp. 145 - 157
• Main Authors: Matta, Sarah Christina, Sankari, Ziad, Rihana, Sandy
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.04.2018
• Subjects: Activity recognition; Artificial neural network; Classification; Confusion matrix; Heart rate variability; ROC; Activity recognition; Artificial neural network; Confusion matrix; Heart rate variability; Classification; ROC
• ISSN: 1746-8094; 1746-8108
• DOI: 10.1016/j.bspc.2018.01.016

•An automatic detection of lifestyle activities is proposed based on Heart Rate Variability analysis using Neural Network Models.•Time- and frequency-domain measures of HRV are good features for discriminating different activity states: sleep, walking, exercising, and eating.•The results have shown that it is possible to classify sleep/awake states based on HRV solely.•The multiclassification of different types of activities showed an accuracy of 88.7%. The quality of life and individual well-being are crucial factors in disease prevention. Particularly, healthy lifestyle lessens the risk and occurrence of main diseases, such as cardiovascular diseases and metabolic disorders. Since a patient has an active role in being a co-producer of his/her health, innovative devices and technologies have been devoted to helping folks in self-evaluation and expected to play a key role to maintain their well-being. In this work, we present a very promising assessment tool for health, Heart Rate Variability (HRV). HRV is the difference in time between one heartbeat and the next. HRV measurement is simple and non-invasive, it is derived from recording of electrocardiogram (ECG) on free-moving subjects. The main aim of this work is to investigate the dynamics in the autonomic regulation of the heart rate by using frequency and temporal analysis to correlate between the HRV and these physiological patterns. In addition to the applied frequency and temporal analyses, pattern recognition is also accomplished using Neural Networks which are further implemented and explored in this work. In the first place, the detection of the sleep/awake states is achieved. Next, a multiclassification of different types of activities such as sleeping, walking, exercising and eating is performed.