Analysis Of EEG Signal Based On Time-varying Parameter First Order Autoregressive Model And Symbolic Transfer Entropy

Electroencephalogram(EEG)is the general reflection of the electrophysiological activity of brain neurons on the cortex or scalp surface.It contains a lot of pathological and physiological information.Therefore,EEG is often used in medical diagnosis and treatment of diseases.In the research and analysis of EEG,the method of nonlinear dynamics plays a very important role.Autoregressive model is a linear model in the analysis of stationary time series,because the EEG signal is a non-stationary time series,autoregressive model are rarely used in the analysis of EEG signals.In 2015,Claus Metzner and Christoph Mark et al proposed first order autoregressive model of time-varying parameters.It can be used to analyze the time series of arbitrary complexity.In this paper,we use this model to analyze EEG signal,and the following is done in this paper:First,this paper uses autoregressive model with time-varying parameter to analyze the EEG signal of epilepsy.Using symmetric channels Fp1 and Fp2 as the input signal,the first-order autoregressive model with time-varying parameter is constructed.Then,using the Bayesian inference method to extract the feature parameter(q _t,a _t),by comparing the peak value of the autocorrelation function of feature parameter a_t,we found the value in epilepsy is always larger than normal person.Therefore,it can be concluded that the first order autoregressive model of time-varying parameters can be used to distinguish between epileptic EEG signal and normal EEG signal.Secondly,this paper use first order autoregressive model with time-varying parameters to analyze sleep stage signal.The channel C4-A1 is used as the input signal,and construct a first order autoregressive model with time-varying parameter for the one-dimensional signal,in this model,we sample the C4-A1 EEG singal,and the sampling interval ist=1,2,3,4,5,6,7,and the experimental result found that it can get best distinction whent=4.Then,using the Bayesian inference method to extract characteristic parameters.By comparing the peak value of the autocorrelation of characteristic parametera_t,we found that the peak value of autocorrelation function parameters of the awake period is always larger than sleep stage I.Therefore,it can be conclude that the first order autoregressive model of time-varying parameters can be used to analyze the sleep stage EEG singal,and can be used to distinguish awake period and sleep I period.Third,this paper uses the symbolic transfer entropy with permutation entropy symbolization to analyze the coupling between awake EEG signal and sleep stage I EEG signal.The experimental results show that the coupling of the EEG singal and ECG signal in awake period is always larger than the sleep stage I.and whatever in awake stage and sleep I stage,the coupling from EEG to ECG is always larger than ECG to EEG.It can concluded that symbolic transfer entropy can be used to analyze the coupling between EEG and ECG in sleep stage signal.