A Machine Learning Model for Automatic Sleep Staging Based on Single-Channel EEG Signals

Abstract

AbstractSleep deeply influences human physiological behaviors. Sleep stage classification is a principal metric for evaluating the quality of sleep. Therefore, it is important to come up with an automatic sleep staging algorithm. The present study presents an automatic sleep staging method that majorly consists of three stages: data processing, feature extraction, and classification. In addition, we presented a proper feature screening technique to discriminate the changes in sleep-level information. Finally, we used the ISRUC-Sleep dataset and raw data from the healthy controlled subjects and subjects with sleep problems to perform model evaluation purposes. The model achieved the highest classification accuracy performance of 98.73% and 96.53% using subgroup-II and subgroup-III respectively, using SVM classification models. In addition, the overall accuracy and kappa coefficient of the proposed method were superior to those of state-of-the-art methods centered on the different machine learning classifiers. In this paper, automatic sleep data staging was realized, effectively improving the accuracy.KeywordsEEGSleep stagesFeature analysisMachine learning