A Deep Neural Model CNN-LSTM Network for Automated Sleep Staging Based on a Single-Channel EEG Signal

Sleep plays a vital role in human physiological behaviors. Sleep staging is a critical criterion for assessing sleep patterns. Therefore, it is essential to develop an automatic sleep staging algorithm. The present study proposes a deep neural network based on a convolutional neural network (CNN) and Long Short-Term Memory (LSTM) for automated sleep stage classification. We presented a deep neural CNN-LSTM network to model character-level information. In the proposed model, the CNN can extract high-level sleep signal features, and LSTM can realize sleep staging with high accuracy by combining the correlations among the sleep data in different sleep periods. Finally, we used the Sleep-EDF dataset for model assessment. On a single EEG channel (Fpz-Oz) from the Sleep-EDF dataset, the overall accuracy achieved 91.12%, according to the results. In most research, the data imbalance of training data exists, which has been solved in the proposed method. In addition, the overall accuracy of the proposed method was superior to those of the latest techniques based on Sleep-EDF. Hence eradicating the tedious work of sleep staging classification required by professionals. The proposed model helped achieve this accuracy level without using any hand-engineered features. The ability of the model to give such conspicuous results without using any handcrafted features makes it quite versatile and robust.