Research On EEG Signal Analysis Method Based On Mode Decomposition

Electroencephalogram(EEG)is a kind of nonlinear complex signal,which contains a large number of information of brain operation and neurological disorders.Human beings'present physiologicalstatus will be learned from effective information extraction based on electroencephalogram.Nowadays,many methods has been used to analyze electroencephalogram,but it is still a lack of preferable analytic method.Empirical Mode Decomposition(EMD)algorithm is an effective method of random signal processing,which can effectively decompose various signals.Therefore,this thesis conducts research based on empirical mode decomposition algorithm for electroencephalogram analysis methods.First of all,in view of the end effect problems in the process of empirical mode decomposition,this thesis presents two kinds of combination and continuation method,which includes support vector machine(SVM)and multiplying windows and least squares support vector machine(LS-SVM)and multiplying window.Simulation experiment and real data experiment show that those two methods above-mentioned can effectively restrain the end effect problems in the process of empirical mode decomposition,and the performance is better than other continuation methods compared.Secondly,choosing the intrinsic mode function(IMF)which is corresponding maximum with original signal to be as feature extraction object of appropriate categories electroencephalogram by correlation analysis,and comparing the characterization performance of several features by the Kruskal-wallis test method,selected a feature of various electroencephalogram by experiments,which are most representative to take as a classifier input,that is average frequency,variance and volatility index.Finally,the thesis comes up with a classifier classification method of electroencephalogram based on empirical mode decomposition algorithm and least squares support vector machine,which is characterized by frequency,average variance and volatility index indicators.Testing the performance of the classification method via doing five experiments and comparing with other different methods.It shows that the proposed classification method based on normal electroencephalogram and epileptic seizure electroencephalogram experiments classified accuracy can be as high as 100%,the comparison of different methods indicate that the classification method has high precision classification performance,and practical effectiveness.At the same time,introducing the least squares support vector machine and multiplying window into the classification method.A large number of experiments shows that empirical mode decomposition algorithm which has been inhibited by end effect can further improve the electroencephalogram classification accuracy.