Methods Of Automatic Sleep Staging And Sleep Assessment Based On EEG Signal

Sleep plays an important role in the body restore and consolidate,with the increase of the pressure of modern life, more and more people suffer from sleep and related illnesses, sleep problems are valued by more and more people. The study of sleep contribute to the prevention and diagnosis of sleep related diseases, to take timely and appropriate treatment, reduce the risk of patients with high-risk disease.The research of sleep staging is the basis of diagnosing sleep related diseases and evaluating sleep quality, and it has important clinical significance.Clinically, sleep staging is generally realized by using manual discriminant method, which is low efficiency and easy cause misjudgment. In recent years, many scholars devoted to automatic sleep staging research, but progress better in abroad, domestic relative lag. The accuracy assessment of sleep quality by using sleep EEG need to be further improved.In this paper, we proposed a new automatic sleep staging system based on single-channel EEG using Least Squares Support Vector Machine(LS-SVM), including the EEG signal acquisition and preprocessing, feature extraction, pattern classification. Based on the automatic sleep staging system, to establish a sleep quality comprehensive evaluation model that combined with the analysis of sleep EEG signal, Pittsburgh sleep quality index and patient self-assessment.It is very difficult to analysis the complicated.EEG signals directly.It is an important task to extract effective sleep features. Based on signal technology and nonlinear dynamics method to extract effective sleep eeg features(10 energy features and 1 Lempel- Ziv complexity). The 10 energy features include total energy, relative energies of K-complex wave, Delta wave, Theta wave, Alpha wave, sleep spindle wave, beta1 wave, beta2 wave, and two band energy ratio( E Ea q, E Ed q). Lempel-Ziv complexity feature can be calculated directly according to its algorithm. Energy feature can be calculated after obtaining the sleep EEG feature waves, comparison of three kinds of EEG feature waves extraction methods: FIR bandpass filter, Hilbert-Huang transform and wavelet packet decomposition, and verify the validity of features.Pattern classification is a key technology in automatic sleep staging system. Based on the principle of pattern recognition, LS-SVM was chosen as the classifier for recognition of sleep EEG features, realize automatic staging finally. Several popular classification method are compared, concluding fisher linear discriminant classifier, LS-SVM classifier and neural network classifier, advantages and disadvantages of each classifier are analyzed and summarized. To design efficient LS-SVM multi-classifier, to test the automatic staging system and validate this automatic sleep staging system.Based on the complexity characteristics of sleep, on the basis of automatic sleep staging system, to establish a sleep quality comprehensive evaluation model, which combined with the analysis of sleep EEG signal,Pittsburgh sleep quality index and patient self-assessment. Scaling method is simple and direct to evaluate sleep, but strong subjectivity. EEG signal detection method is an objective method, but need to a large number of signal processing steps, and the evaluate result is limited by staging accuracy rate. Considering the advantages and disadvantages of various methods, the establishment of sleep quality comprehensive evaluation system can make the sleep assessment more objective and accurate.