Research Of Automatic Sleep Staging Method Based On EEG Signals

Sleep is the most important physiological process related to a person's physical health.With the development of society,the acceleration of people's living rhythms and the increasing pressure on life have led to more and more people suffering from reduced sleep quality,and the diseases derived from them have also increased.Therefore,the research of sleep has attracted widespread attention from scholars.Clinically,by staging the sleep process to detect related diseases,the traditional artificial sleep staging methods are time-consuming,laborious,and subjective.Electroencephalography(EEG)is an important basis for the diagnosis and treatment of sleep disorders.In recent years,research on the automatic staging method of sleep EEG signals has important application value for the treatment and diagnosis of sleep disorders.This paper studies a new method of automatic sleep staging based on EEG signals,which mainly includes three parts: signal preprocessing,feature extraction and classification recognition.Firstly,the EEG signals of 8 testers in the MIT-BIH database were used as the analysis objects,the wavelet transform was used to preprocess the EEG signals,and the wavelet packet decomposition method was used to extract the four rhythm waves of the EEG signals,namely ?,?,?,and ?.And calculated the relative energy characteristics,and then extracted the complexity of EEG,multi-scale sample entropy(MSE),Lyapunov exponent and correlation dimension features.In this paper,the MSE method was improved,and a new method of compound multi-scale sample entropy(CMSE)was proposed,which reduced the calculation errors and improved the calculation ability.Construct two classification models of BP neural network and support vector machine(SVM)for all the characteristic parameters for automatic staging,and compared the staging results with the expert staging to get the accuracy rate,choose a more accurate SVM classifier for analysis of experimental data.Finally,the 48-hour sleep EEG data collected by the hospital sleep acquisition module for eight patients was used to extract multiple features,using MSE and CMSE as the main features,and the SVM method was used for stage calculation,and the automatic analysis results were compared with the hospital diagnosis staging results,which verified the high accuracy of SVM sleep staging with CMSE as the main feature.The experimental results show that both BP neural network and SVM can be used as models for sleep staging.The overall staging results are similar,but the accuracy ofautomatic staging by SVM is higher.When using SVM to perform automatic staging of sleep features such as CMSE,rhythm waves,relative energy,complexity and other characteristics proposed in this paper,it can reduce the errors,the calculation time of feature extraction,and improve the accuracy of automatic staging.Therefore,the sleep staging method based on EEG signals proposed in this paper provides an effective way for the diagnosis and treatment of sleep disorders,which has practical application value.