A Virtual Mouse Based on Parallel Cooperation of Eye Tracker and Motor Imagery.

It has become a reality for paralyzed patients to use independent BMI-based assistive application to facilitate their lives. However, it was still a problem to switch among applications, if applications like typing and wheelchair-control were integrated in a device. In order to solve this problem, this paper realized a virtual mouse by combining eye tracker and MI, hoping to help patients to switch BMI applications independently and conveniently. The improved VT filter proposed in this paper solved the stable validity problem of the cursor. During the 2000 ms for MI, the time when the mouse stabilized in the valid area of the file can reach 94%. Asynchronous MI was realized in this paper through sliding window to enhance the flexibility of the system. The classification accuracy of left hand and right hand was 92.36%, that of left hand and idle state was 90.28%, and that of right hand and idle state was 90.63%. The final results were obtained by voting, which reached 91.33%. The results proved that the decoding of MI data would not be impacted during the control of eye tracker, which meant that the parallel control mechanism toward multi-cognitive modality proposed in this paper was feasible. This multi-cognitive modality based parallel control mechanism is in line with cognitive habits of human beings, and promising in improving the performance of human-machine fusion systems, in the future.