Improved Deep Neural Network for Automatic Staging Applied to Assisted Sleep System
Download as PDF
Meng Lu, Jiawen Chen, Shengpeng Shi and Zhenyu Zhu
In this paper, aiming at sleep problem, we propose an improved automatic sleep scoring model based on convolutional neural network and time-distributed Long Short-Term Memory. This model is applied to real-time assisted sleep system for the intelligent wearable device with client-server architecture in mobile application and website. In this model, we employ CNN to extract time-invariant features from raw EEG epochs, employ LSTM to learn temporal information from a sequence of EEG epochs that have been extracted features, and utilize time-distributed layers to connect the two parts. Sleep EEG signals from open Sleep-EDF database are used to evaluate the performance of our proposed model. The results show that our proposed model has good classifications for most sleep stages, especially for awake and SWS stages. Moreover, we obtain overall accuracy of 85.7%, macro F1-score of 80.5%, and Cohen’s Kappa coefficient (κ) of 0.81 on 20-fold cross-validation for five sleep stages according to AASM, which outperforms other six existing methods. Due to better staging performance of our model, our assisted sleep system can have a better adaptive adjustment for people and it also has a good market promotion prospect and potential commercial value.
Sleep Stage Classification, Deep Neural Network, Assisted Sleep System, EEG Signals