Abstract
Noninvasive electroencephalogram (EEG)-based brain-machine interface (BMI) provides a new communication channel to replace an impaired motor function of patients. EEG recordings during right and left motor hand imageries are expected to be available to move a cursor on a computer screen or to move a wheelchair. In the present experiments, visual cues ("right hand" or "left hand") were presented. The subjects were asked to move one of their hands or imagine it when visual cues were presented. First, blinking artifacts were eliminated by using ICA. Secondly, spline Laplacian analysis was applied. Thirdly, the most significant channel, latency, and frequency band were determined for the discrimination based on time-frequency analyses with wavelets. Lastly, in their parameters, trials were discriminated between right and left hand movements or their imageries. The highest percent correct of the discrimination in the 5 subjects was over 70%, indicating the presented method may be applicable as a BMI.
