Motor imagery multi-task classification method

Abstract

We studied on the MI-BCI (Motor Imagery Brain-Computer Interface). MI-BCI is an interface that operate a computer using changes in brain activity that appear when imaging moving a body part. For example, MI-BCI is possible to assign the left-hand motor imagery to the power ON/OFF command. A problem of MI-BCI is a few number of the command. Currently, MI-BCI commands are four commands using "left-hand", "right-hand", "legs" and "tongue" motor imagery. Therefore, we attempted to add the number of MI-BCI commands by classifying eight kinds of motor imagery brain activity "no movement", "left-hand", "right-hand", "legs", "both-hands", "left-hand + legs", "right-hand + legs", "both-hands + legs". Motor imagery by multiple body parts "both-hands", "left-hand + legs", "right-hand + legs", "both-hands + legs" are called multi-task. Multi-task are combination of simultaneous motor imagery of left-hand, right-hand, and legs. This makes it possible to add the number of commands to 2N − 1 (N is number of body part). We used LDA to classify motor images. As a result of classification, the correct classification rate was 26.9%. It was shown that multitask motion recall can be classified, and it was suggested that it is possible to add the number of MI-BCI commands to2N − 1.