Sleep Staging Algorithm And Application Research Based On Single Channel EEG Signal

Sleep staging is the division to the progress people experienced during sleep.Accurately sleep staging monitoring can provide technical support for sleep disease diagnosis,and then take appropriate regulate and treatment.On one hand,the sleeping state transition process and potential problems could be detected accurately from sleep staging.On the other hand,continuous monitoring can effectively prevent sleep breathing heartbeat pause in the sudden illness,and protect human life and health.The polysomnography sleep staging method is commonly used in hospital.However,due to its much electrodes,complex measuring method,greatly influence to the monitoring object and easily induce physical artifact,the development and application is seriously limited.Portable or wearable sleep staging monitoring equipment with less electrodes has become one of the hot research areas.Several key techniques sleep staging from single channel EEG signal are researched.The main contents are as follows:(1)Single channel EOG artifact removal.The EOG artifacts generated by eye movements in REM stage is the major artifact in EEG.The independent component analysis(ICA)has been applied to multi channel EOG artifacts removal successfully.But,in single channel channel,the WT-ICA artifacts removal method has the problem of overcomplete and unstable.In this article,a novel artifacts removal method WT-EEMD-ICA algorithm used in single channel EEG artifacts removal is proposed by adding the EEMD algorithm.The overcomplete problem in WT-ICA is successfully solved.In the condition without reference EOG channel,the EEG and EOG artifacts are successfully separated only from one single channel EEG signal.(2)Single channel EEG feature extraction.The traditional EEG feature extraction in multi-channel is usually focused on single domain analysis.The variation and distribution of parameters from each channel are extracted.However,due to the complexity of EEG,it is difficult to extract enough information from single channel sleep EEG only from one single domain parameters.A novel feature extraction method for single channel EEG is proposed based on the multi-domain analysis of the EEG in this article.The multifractal detrended fluctuation index,the visibility graph of series,the fast Fourier transform subband energy ratio and short time Fourier transform distribution of parameters are extracted from the analysis of nonlinear,time domain,frequency domain and time-frequency.The sleep staging effective of the extracted parameters are verified compared to the ApEn,L-Z complexity,SyEn and AR model.(3)Single channel EEG feature selection.It is necessary to optimize all the extracted parameters to establish a excellent sleep EEG model with high stability and generalization ability.In the commonly used simulated annealing genetic algorithm,the new solution generated from neighborhood randomly will bring serious impact on the stability of the solution,even no convergence about the iteration results.In this article,an adaptive simulated annealing genetic algorithm(ASAGA)is proposed by adding the adaptive crossover and mutation probability adjust mechanism,the genetic optimization selection mechanism,and the design of weighted fitness function.The effectiveness of ASAGA is demonstrated compared to the other four algorithms.(4)Single channel EEG feature classification.Due to the highly non-stationary and nonlinear characteristics of the EEG,the classification accuracy of nonlinear methods are higher than the linear methods.The adaptive network based fuzzy inference system(ANFIS)is the typical one which widely used in classification.But,the classification accuracy of ANFIS is usually not excellent since the rule base should be created by the experience.In this article,an optimal GA-ANFIS is proposed by adding the genetic algorithm(GA)to the establishment of rule base,and the classification accuracy is greatly improved compared to the PLS?LS-SVM and ANFIS.(5)Experimental verification.The effective extraction of EEG is the foundation of sleep staging monitoring,the EEG collected from scalp is a low frequency weak signal with 10?100?V and 0.5?100Hz,and the background noise and interference are also very serious.Based on the above characteristics of EEG,a portable sleep monitoring equipment with single channel EEG that has three electrodes is designed.The CMRR is higher than 100 dB and the input impedance is higher than 50M?.A joint sleep staging monitoring experiment system is constructed by using the portable sleep staing monitoring equipment and IoC-View which designed by Morpheus Medical,and the results demonstrated the effectiveness of the equipment and software.