Feature selection strategy for classification of single-trial EEG elicited by motor imagery

Brain-Computer Interface (BCI) provides new means of communication for people with motor disabilities by utilizing electroencephalographic activity. Selection of features from Electroencephalogram (EEG) signals for classification plays a key part in the development of BCI systems. In this paper, we present a feature selection strategy consisting of channel selection by fisher ratio analysis in the frequency domain and time segment selection by visual inspection in time domain. The proposed strategy achieves an absolute improvement of 7.5% in the misclassification rate as compared with the baseline system that uses wavelet coefficients as features and support vector machine (SVM) as classifier.