Feature selection and classification on brain computer interface (BCI) data

In this paper, a large number of features are extracted from raw EEG data and then feature selection and classification are performed ,for brain computer interface (BCI) applications using motor imaginary movements. As the feature selection method, mRMR (minimum Redundancy Maximum Relevance) method, which is a fast method to select relevant and non redundant feature set, is chosen. Using a number of different classifiers, it is observed that feature selection helps with the classification performance, higher classification accuracy is achieved using less features. In the experiments, the BCI Competition 2003 3A data set is used.