Classification of Motor Imagery Based on Multi-Decision Fusion for Brain Computer Interface

Brain-computer interface (BCI) based on motor imagery (MI) is considered to be a promising cognitive tool for rehabilitation therapy of movement disorders. Feature extraction and classification are important in MI-BCI, and affect the BCI's performance. Label is the classifying result of motor imagery in MI-BCI. For example, 1 represents imaging left hand movement and 2 represents imaging right hand movement in a MI-BCI system. In this paper, we combined labels from different feature extraction and classification, and fused labels to get a new label which has higher accuracy. We named the method multi-decision fusion. We used the multi-decision fusion on BCI competition's MI datasets for classification. By comparing the results with those using conventional methods of feature extraction and classification, we have verified multi-decision fusion is an effective method. Multi-decision fusion can effectively improve classification accuracy of motor imagery.