Feature Extraction Of Eeg And Sleep Stages Were Studied

Information of EEG is one of the most important clues for diagnosis of mental condition. Besides of being diagnostic tool, EEG detection and processing hold the balance in Brain-Computer Interface systerm. Recently, sleep disorders is becoming the killer of human health. More and more people are feazed by sleep disease. The diagnoses of sleep diseases, which contain insomnia, hypersomnia and so on, are the research focus.This dissertation mainly focuses on the feature extraction of EEG and sleep stage processing. First of all, the paper depicts the theory of EEG, physiological characteristic, the methods of EEG detection and the theory of polysomnography. The chapter 3 summarizes the time-frequency method, Neural Network and so on in EEG processing. Particularly Wavelet Transform (WT) is used in eliminating noise and feature extraction. Also, nonlinear dynamic and ICA are discussed. The FastICA algorithm is used in EEG basic feature extraction. The algorithm is limited because the transcendental information of EEG must be realized. Then the constrained ICA is introduced in EEG feature extraction. According to PSG, seventeen parameters are calculated for sleep segmentation processing. The improved BP network, which can eliminate the limitation of input data order, is used for sleep segmentation. The emulational results show that the nicety percent of total sleep stages is reached 79.2%. The dissertation predicts the future trend of EEG processing and sleeps stages in the end.