Research Of Automatic Sleep Staging Based On EEG Signals

With the accelerated pace of social life, the pressure on all of social life in all aspects is increasing, and the impact of sleep diseases on people’s lives is growing. Where insomnia is the most common sleep disorders affect people’s lives diseases. Electroencephalogram (EEG) is an important basis for the diagnosis and treatment of sleep disorders. EEG reflects the brain is the main cell population electrophysiological activity. When people enter deep sleep, the brain nerve cell group slows down the activity, and the EEG signal also reflects this change. Due to the EEG signal non stationarity, stochastic, nonlinear, low signal to noise ratio characteristics, so the effective sleep staging has been difficult to study, low efficiency of the traditional artificial sleep staging, time-consuming and laborious, so sleep EEG automatic staging to improve sleep quality, medical clinical research has important significance.The main purpose of this paper is to study a method of automatic staging of EEG signals based on two lead. Experimental data from the Sleep-EDF MTT-BIH database of 8 normal subjects sleep EEG. In this paper, the design of EEG automatic staging system is realized from three aspects:signal preprocessing, EEG feature extraction and feature parameter classification. The main contents of this paper are as follows:1. Method using wavelet transform method for pretreatment of sleep EEG of the original signal, using wavelet packet to extract EEG basic rhythms wave alpha, theta, delta waves, beta waves and calculated the alpha wave, theta, delta waves and beta Potter syndrome relative wave energy (Ej/Eall), the total energy of the sleep EEG signal value Eall and alpha waves and theta waves of energy ratio Ea/Ee, delta waves and theta wave energy ratio Eδ/Eθ.2. Firstly sample entropy extraction algorithm to extract the sample with characteristic EEG entropy, and the feasibility of fuzzy entropy as the feature parameter of sleep stage is studied in view of the shortcoming of template matching. The extracted basic rhythm delta waves and theta wave as the feature vector selection method of fuzzy entropy of the various stages of sleep were analyzed, and finally extracted sleep EEG fuzzy entropy fuzzy entropy feature.3. On the basis of the extracted signal characteristics utilizing SVM for EEG sleep stages automatically verify classifier generalization ability. By using the energy characteristic value, the characteristic value of the sample entropy and the characteristic value of the fuzzy entropy, the EEG signal is staged and the experimental results are analyzed. According to the characteristics of EEG and combined with entropy characteristic properties, energy characteristics, is proposed based on sample entropy, fuzzy entropy, energy characteristic values combination of sleep EEG automatic staging method and verify the method of classification results.The main innovation of this paper are listed as follows:1. Based on Sample Entropy algorithm must contain template matching, fuzzy entropy feasibility study sleep stage characterized by parameters. The basic rhythm delta waves and theta wave as the feature vector extraction, fuzzy entropy method to carry on the analysis of the various stages of sleep and verify the fuzzy entropy variation tendency in each stage of sleep.2. In view of the characteristics of EEG signal, energy feature and characteristics of the entropy, energy feature is proposed in this paper, sample entropy, fuzzy entropy feature combination way of sleep EEG automatic staging.Experimental results show that the wavelet packet, sample entropy, fuzzy entropy can effectively extract the sleep characteristics of sleep EEG. In the use of characteristics of sleep sleep stage of wavelet packet energy feature extraction of recognition correct rate is low, sample entropy feature recognition correct rate relative to the energy feature is obviously improved and fuzzy entropy feature in sleep EEG stage relative to the sample entropy method and increased about 5%. The proposed energy characteristics, features sample entropy, fuzzy entropy feature a combination of sleep EEG performed automatically staging, and staging accuracy is higher than that of single energy eigenvalues and entropy feature.