A Study on Automatic Sleep Stage Classification Based on CNN-LSTM

Automatic Sleep Stage Classification (ASSC) plays an important role in the diagnosis of sleep related diseases. However, due to the complexity of mathematical modelling, ASSC has many difficulties. At the same time, the rapid fluctuations between the adjacent sleep stages make it difficult to extract features, resulting in an inaccurate classification of a period of electroencephalogram (EEG) sleep. In order to solve the above problems, this paper proposes a sleep stage classification method based on convolutional neural network and long-term short-term memory network (CNN-LSTM). The method applies CNN to extract spatial features from the original data and LSTM to extract temporal features and adopt softmax to classify these features. To verify the proposed method, we tested it on a public data set called ISRUC-Sleep and compared it with several state-of-the-art methods. The experimental results show that the proposed method significantly improves the accuracy of sleep staging and achieves better results.