A Deep Learning-Based Method for Sleep Stage Classification Using Physiological Signal

Guanjie Huang, Chao-Hsien Chu, Xiaodan Wu

Abstract: A huge number of people suffers from different types of sleep disorders, such as insomnia, narcolepsy, and apnea. A correct classification of their sleep stage is a prerequisite and essential step to effectively diagnose and treat their sleep disorders. Sleep stages are often scored by experts through manually inspecting the patients’ polysomnography which are usually needed to be collected in hospitals. It is very laborious for experts and discommodious for patients to go through the process. Accordingly, current studies focused on automatically identifying the sleep stages and nearly all of them need to use hand-crafted features to achieve a decent performance. However, the extraction and selection of these features are time-consuming and require domain knowledge. In this study, we adopt and present a deep learning approach for automatic sleep stage classification using physiological signal. Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) of popular deep learning models are employed to automatically learn features from raw physiological signals and identify the sleep stages. Our experiments shown that the proposed deep learning-based method has better performance than previous work. Hence, it can be a promising tool for patients and doctors to monitor the sleep condition and diagnose the sleep disorder timely.