| Sleep occupies almost one-third of human’s life.Insufficient sleep can induce sleep disorders,neurasthenia and,other sleep diseases.Therefore,the assessment of sleep stage is of great significance for doctors to grasp the patient’s sleep status.Compared with other physiological signals,non-invasive electroencephalogram(EEG)signals can most directly reflect changes in the physiological system and are widely used in sleep analysis.Traditional sleep EEG analysis often adopt linear analysis methods based on time domain,frequency domain,and time-frequency domain.EEG is a nonlinear and non-stationary signal.Linear analysis often ignore some nonlinear characteristics of EEG signals.Therefore,this paper adopted two types of non-linear methods,which is the fractal-based methods and the entropy-based methods,to reveal the potential mechanism behind sleep EEG signals in different views.During the process of nonlinear analysis,this paper puts forward the enhanced method to hand out the problem based on the traditional methods.The main research contents of this paper are as follows:The first part adopted the Wavelet leader multifractal formalism(WLMF)to analyze the complexity of the Wake stage and the S3 stage of the sleep EEG.The multifractal spectrum calculated by the WLMF algorithm can reflect the multifractal characteristics of the dataset from two aspects(the spectral width and the holder exponent).The dataset was divided into different groups according to 3 factors(the age,the gender,and the brain position).The influence of multifractal spectra(the Wake stage and the S3 stage)by different factors are studied.The second part adopted another complexity algorithm,multiscale entropy(MSE),to analyze the complexity of the Wake stage and the S3 stage of the sleep EEG.Based on the research in the previous part,only the influence of aging on the complexity of the Wake stage and the S3 stage were studied in this part.The result shows that the MSE curve of the Wake stage and the S3 stage appeared cross-over phenomenon in both the young group and the elderly group.It’s clearly that the complexity of the Wake stage is greater than that of the S3 stage on the smaller scale,while the complexity of the Wake stage is less than that of the S3 stage on the larger scale.This cross-over phenomenon is inconsistent with the conventional physiological cognition.The third part focused on the research of the cross-over phenomenon in the MSE curve of the Wake stage and the S3 stage.Firstly,the simulated experiment was designed to reveal the cause of the cross-over phenomenon.It proved that the EEG’s complexity based on multiscale entropy is easily affected by the rhythmic signals.Therefore,we proposed a multiscale entropy analysis of instantaneous frequency variation(MSE-IFV)method based on the multiscale entropy,and adopted the real-world sleep dataset for algorithm’s verification.The results show that MSE-IFV can effectively improve and solve the cross-over phenomenon.This paper adopts two types of nonlinear complexity estimated methods to analyze EEG signals.The nonlinear characteristics of sleep EEG signals was analyzed from different aspects.Experimental results show that aging can affect the distribution of multifractal spectra of the Wake stage and the S3 stage.This finding broadens the application of multifractal spectra in the analysis of the sleep EEG.The proposed MSE-IFV method effectively reveals and solves the cross-over phenomenon of the Wake stage and the S3 stage in the MSE’s curves,and expands the application of the nonlinear method based on MSE. |
