Combined feature extraction method for classification of EEG signals
- 540 Downloads
Classification of electroencephalogram (EEG) signals is an important task in brain–computer interfaces applications. This paper combines autoregressive (AR) model and sample entropy and presents a combination strategy of feature extraction. Each feature vector obtained from the combination strategy contains two parts: AR coefficients and sample entropy values. In the classification phase, this paper employs support vector machine (SVM) with RBF kernel as the classifier. The proposed method is used in the five mental task experiments. Experimental results show that the SVM classifier performs very well in classifying EEG signals using the combination strategy of feature extraction. It obtains a better accuracy in comparison with AR-based method. The results also indicate that the combination strategy of AR model and sample entropy can effectively improve the classification performance of EEG signals.
KeywordsEEG signal Feature extraction Autoregressive model Sample entropy Support vector classification Brain–computer interfaces
This work is partly supported by National Natural Science Foundation of China (No. 61373127) and the University Scientific Research Project of Liaoning Education Department of China (No. 2011186).
- 5.Richman J, Moorman J (2000) Physiological time series analysis using approximate entropy and sample entropy. Am J Physiol 278(6):2039–2049Google Scholar
- 8.Ianez E, Azorin JM, Ubeda A, Fernandez E, Sirvent JL (2010) LDA-based classifiers for a mental tasks-based brain–computer interface. In: Proceeding of the 2010 IEEE international conference on systems man and cybernetics (SMC 2010), 10–13 Oct 2010. IEEE Press, pp 546–551Google Scholar