The Analysis And Classification Of Steady-state Visual Evoked Potential And Schizophrenia MEG Signal

EEG/MEG are complex biological signal that reflect the different physiological state of the brain,commonly used in diagnosing and monitoring of epilepsy,alzheimer disease,schizophrenia and other diseases.The analysis of the pathological EEG/MEG signal is benefit to further understand underlying mechanism of brain disease,and provide reference for clinical diagnosis,and make the patient be more hopeful.This thesis studies the MEG signal of schizophrenia and the EEG signal of SSVEP.In this thesis,the analysis methods of EEG/MEG signal are introduced.It focuses on feature extraction and pattern classification methods,the principle and characteristic of these methods are described in detail.An approach based on multidimensional complexity features analysis for MEG signal was proposed in this paper.Several features including the AR model coefficients,frequency band energy,approximate entropy and Lempel-Ziv complexity are extracted.The distance criterion and plus-L minus-R algorithm are used to filter the channel,then BP neural network and SVM are employed to classify schizophrenia and control participants signal,classification accuracy are 96.25% and 98.75%.The result shows that the proposed method can distinguish schizophrenic patients and control participants effectively.The thesis also uses the genetic algorithm to select the features with significant differences.A classification accuracy of 98.5% and 99.75% is obtained by BP Neural Networks and SVM respectively,and the SVM can obtain better classification performance.Finally,the EEG experiment based on SSVEP is designed,and acquired the EEG signal of three subjects.The DFT,CCA and MSI methods are used to analyze the EEG signals.The DFT spectrum analysis shows that the signal energy is largest at the target stimulus frequency.The correlation coefficient and synchronization index is largest at the target frequency in CCA and MSI method.The frequency recognition result of CCA method is superior to MSI and DFT,whether in short or long time window.MSI method performance is generally better than DFT,especially in short time window.