Comparison of CNN-Uni-LSTM and CNN-Bi-LSTM based on single-channel EEG for sleep staging

Abstract

Sleep staging is an effective method for diagnosing sleep disorder and monitoring sleep quality. With the rapid development of machine learning technology, the automatic staging methods of sleep gradually replace the traditional manual interpretation which can improve the efficiency on sleep staging for medical research. LSTM networks can save the historical information as a reference for the current moment, which is undoubtedly a good way to improve sleep staging performance. In this paper, a convolutional neural network (CNN) is constructed to extract the features from a single-channel EEG. The Uni-directional Long Short-Term Memory (Uni-LSTM) network and Bi-directional Long Short-Term Memory (Bi-LSTM) network are combined with CNN to realize automatic sleep staging. The obtained results showed that the two presented network frameworks are effective and feasible on sleep staging. The Bi-LSTM which has more enriched sequence information got better classification performance than the Uni-LSTM.