Research On EEG Signal Recognition Based On Variational Bayesian Depression

Depression is a common mental illness.It is characterized by low mood and pessimism.Suicidal behavior may occur when symptoms are severe.As the number of depressive patients increases year by year,the diagnosis results are subject to subjective factors,which can easily lead to misdiagnosis and missed diagnosis.Therefore,it is urgent to improve the accuracy of its diagnosis.Through the comprehensive analysis of the research status,processing and analysis methods of electroencephalogram(EEG)in depression,it is found that the selection of EEG denoising method and diagnostic model is very important to improve the accuracy of diagnosis.In terms of denoising,this paper proposes a Hilbert-Huang Transform(HHT)combined with wavelet packet denoising method.The collected EEG signal contains noise,and the noise-containing Intrinsic Mode Function(IMF)component obtained in the HHT is further processed by the wavelet packet algorithm,so that the denoised reconstruction signal can retain more effective information.In the establishment of the diagnostic model,due to the great uncertainty in the EEG signals,the use of a probability-based representation method can often achieve good results for the analysis of EEG signals.Therefore,this paper establishes a hidden Markov diagnosis model of EEG based on the variational Bayesian theory.First,using the variational Bayesian Juliu(VB)algorithm,the maximum likelihood(ML)algorithm and the maximum posterior(MAP)algorithm to estimate and analyze the parameters of the hidden Markov model(HMM).Then,the hidden Markov model obtained by the VB and ML algorithms is compared on the synthetic data.The experimental results show that the VB algorithm can avoid the over-fitting phenomenon in the ML algorithm.Finally,the fluctuation index of the denoised EEG signal is used as the input of the diagnostic model,and the recognition rate of the hidden Markov model under the three algorithms is compared.The experimental results show that the hidden Markov model under the VB algorithm has the highest recognition rate for depressive and healthy controls,reaching 91.2%.