SLEEP STAGES CLASSIFICATION BASED ON BIORADAR USING CNN-LSTM NEURAL NETWORK

The research is devoted to the automatic sleep stages classification based on non-invasive biosensor. A deep learning framework for classification of wakefulness, REM and NREM sleep stages (N1, N2, and N3) is proposed. The system consists of a combined convolutional and recurrent (long short-term memory, LSTM) neural network (CNN-LSTM). The proposed CNN-LSTM neural network significantly outperforms existing machine learning methods based on traditional manual feature engineering. The model achieves an accuracy of 0.88 and a Cohen's Kappa agreement coefficient of 0.84, that is almost perfect agreement. The results of the study could be a promising solution for automatic sleep assessment without manual data processing and can be very useful for sleep screening.