The Analysis And Automatic Staging Of Sleep EEG

Everyone's life is about one-third of the time in sleep,and sleep is an essential physiological activity.Through sleep,people's physical strength can be supplemented,the spirit can be restored,and good sleep is beneficial to physical and mental health.However,with the acceleration of the pace of social life,work and study pressures have increased,and sleep-related diseases have seriously affected human health.Effective and objective assessment of human sleep quality is conducive to the prevention and treatment of sleep disorders.One of the important methods for assessing sleep is based on sleep electroencephalogram(EEG)signals for sleep staging.This is also the prerequisite for an objective assessment of sleep quality.The EEG signal reflects and records the state of the brain's physiological activities,and studies the characteristics of EEG signals.It is the basis for studying sleep stages,improving sleep quality,and diagnosing sleep diseases.It has important theoretical significance and application value.The main content of this paper is to study a method based on single-channel sleep EEG signal for automatic sleep staging.The data from this article is derived from Sleep EDF database sleep signal in Physio Bank.The main contents are denoising of sleep EEG signals,extraction of rhythm waves,extraction of sleep characteristics,and automatic staging of sleep.The main contents include:(1)The wavelet threshold denoising method for signal denoising is used to perform5-layer wavelet decomposition through the original EEG signal,and the soft thresholding method is used to complete the denoising processing of the EEG signal;(2)Using 7-level wavelet decomposition and 6-layer wavelet packet decomposition and reconstruction respectively,the rhythm wave is extracted and parallelly compared.Finally,the wavelet packet is used to complete the extraction of the sleep rhythm wave.(3)Using the sample entropy to extract the entropy features of the 4 kinds of rhythmical waves of EEG signals under different sleep states.In addition,the entropy of the 9 samples,10 samples,11 samples,12 samples,and 13 samples are used for the denoised EEG signals.Sleep EEG feature extraction;(4)Using the above 9 features as the input of the classifier and the random forest and support vector machine as the classifier,the conclusion is drawn that the use ofsupport vector machines is more suitable for the EEG sleep staging.The experimental results show that the sample entropy of the rhythm wave and the multiscale entropy of the de-noising sleep EEG signal are all effective features for sleep staging.In the processing,the above 9 classification features are used as input,and the data amount is about 12,000 for a 30 s sleep signal.At the time,the support vector machine was more accurate than the random forest staging result.