Sleep stage classification using the combination of SVM and PSO

Abstract

Sleep quality is one of the most important factors for human physical and mental health. Sleep disorder may increase the risk of developing chronic physical and mental illnesses such as heart failure, coronary heart disease, depression, and bipolar disorder. In addition, sleep disorder also decreases work productivity and increases the risk of traffic accidents. The problem of sleep disorder is usually associated with the irregularity in sleep cycles. People need to get the right proportion of every stages and sufficient number of cycles to obtain a quality sleep. The aim of this study is to examine distinctive features related to sleep stages (wake, light sleep, deep sleep) from heart rate variability (HRV), and evaluate their usefulness to classify sleep stages. We utilize support vector machine (SVM) to classify the sleep stages classification and compare the result with conventional methods. We also utilize particle swarm optimization (PSO) for feature selection. The simulation results show that our proposed sleep classification with SVM and PSO can improve the accuracy of sleep stage classification.