The precise classification for the electroencephalogram(EEG)in different mental tasks in the research on braincomputer interface(BCI)is the key for the design and clinical application of the system.In this paper,a ne...The precise classification for the electroencephalogram(EEG)in different mental tasks in the research on braincomputer interface(BCI)is the key for the design and clinical application of the system.In this paper,a new combination classification algorithm is presented and tested using the EEG data of right and left motor imagery experiments.First,to eliminate the low frequency noise in the original EEGs,the signals were decomposed by empirical mode decomposition(EMD)and then the optimal kernel parameters for support vector machine(SVM)were determined,the energy features of thefirst three intrinsic mode functions(IMFs)of every signal were extracted and used as input vectors of the employed SVM.The output of the SVM will be classification result for different mental task EEG signals.The study shows that mean identification rate of the proposed algorithm is 95%,which is much better than the present traditional algorithms.展开更多
基金This work is supported by National Natural Science Foundation of China under Grant No.81071221.
文摘The precise classification for the electroencephalogram(EEG)in different mental tasks in the research on braincomputer interface(BCI)is the key for the design and clinical application of the system.In this paper,a new combination classification algorithm is presented and tested using the EEG data of right and left motor imagery experiments.First,to eliminate the low frequency noise in the original EEGs,the signals were decomposed by empirical mode decomposition(EMD)and then the optimal kernel parameters for support vector machine(SVM)were determined,the energy features of thefirst three intrinsic mode functions(IMFs)of every signal were extracted and used as input vectors of the employed SVM.The output of the SVM will be classification result for different mental task EEG signals.The study shows that mean identification rate of the proposed algorithm is 95%,which is much better than the present traditional algorithms.
