EEG based method for the decoding of complex arm motor imagery tasks

Brain-computer interface (BCI) is a new kind of communication and control technology, which connect the human brain to external world by converting users' intention into machine command without the cooperation of normal nerves and muscles. Recently, brain computer interface based on motor imagery (MI) has received increasing interest for its practicability and convenience. However, the short of imagery pattern makes the application of MI difficult to realize. This paper will propose a MI patterns including four novel and complex arm gestures: clockwise and anticlockwise swing of both arms. Preliminary result shows that using support vector machine classifier and deep brain network classifier, we are able to discriminate these tasks with average classification accuracy of 54.41%, and average information transmission rate of 8.05 bits/min. Meanwhile, the result shows clockwise and anticlockwise movements of same arm (error rate: 17.33%) are not as easily to discriminate as movement of left and right arms (error rate: 15.91%).