A Brain Computer Interface Based on Motor Imagery for Maze Game

Motor imagery electroencephalograph (EEG) signals can be applied in brain computer interface (BCI) systems to make up specific control instructions. In this paper, the common space pattern (CSP) algorithm is used for spatial feature extraction of four different motor imagery (MI) tasks as well as sorting and classifying the features by Multi-Cluster Feature Selection (MCFS) algorithm and Sparse Representation Discrimination Analysis (SRDA) algorithm, the average classification rate can reach 68%. In order to emphasize the feasibility of MI EEG in non-invasive BCI, a maze game is designed and implemented. Subjects can control the process of game successfully in an unshielded environment.