Automatic Sleep Stage Scoring Using Hilbert-Huang Transform with BP Neural Network

In this paper, a novel method based on Hilbert-Huang Transform (HHT) and backpropagation (BP) neural network is proposed to perform automatic sleep stages classification. Features extracted from 30-second epoch of EEG using HHT are good representations of EEG signal. A three-layer BP neural network is employed to classify these features to one appropriate stage. For a four-stage classification, consisting of Awake, Stage 1 + REM, Stage 2 and slow wave stage (SWS), of one single Pz-Oz channel EEG signal alone from 7 human subjects, the average stage recognition rate of the proposed method can achieve Awake 95.2%, Stage 1+Rem 87.1%, Stage 2 82.0%, SWS 92.9%. The experiment results show the method is effective and promising in automatic sleep states classification. It can be a powerful tool in sleep quality monitoring and sleep-related diseases diagnosis.