Comparison between Effective Features Used for the Bayesian and the SVM Classifiers in BCI

Brain-computer interface (BCI) is based on processing signals recorded from the scalp, the surface of the cortex or from the inside of the brain in order to identify desired actions or behaviors. In BCI we are interested in extracting the most effective features from rare data in order to have the desired classification results. In this paper besides proposing two discrimination algorithms for classifying imagined movements of the left small finger and the tongue, a comparison has been done between the effective features applied by the Bayesian and the SVM classifiers for the BCI task. In fact the comparison was done on the most effective features found from a pool of extracted features for each classifier, separately. Finally using the most effective features of each classifier, the classification accuracy of 89.21% and 91.01% were achieved for the Bayesian and the SVM classifiers, respectively.