Ensemble methods combination for Motor Imagery tasks in Brain Computer Interface

The correct recognition of Motor Imagery task in Brain-Computer Interface (BCI) systems has been an important issue in recent studies. In this study, we propose a classification framework based on ensemble methods to handle spectral and spatial EEG signal characteristics. A mixture of two ensemble classifiers has been used for combining multiple information sources. The performance of the proposed classifier has been evaluated on a two-class problem (right and left hand) from the BCI Competition IV dataset 2a. The used features for the training data are the selected features by Mutual information-based Best Individual Feature from the output of the Filter Bank Common Spatial Pattern. The results show that proposed method can reach an accuracy of 90.27% with just 7 features, while other methods have lower accuracy and a higher number of features.