The accuracy on electroencephalogram (EEG) based motor imagery recognization of different movements of a same limb is low. It constrains the monitor method

which monitor motor imagery training of patients with limb paralysis based on brain computer interface

to be used in clinical. Aiming at the problem

in this paper a new method of multiscale recognization of motor imagery is proposed. It is based on the coupling relationship between eye movement and imagined movement during motor imagery. In this method

first of all EEG is applied to judge whether motor imagery occurs on a large scale

and then combined with the recognization result of electrooculogram (EOG)

which changes cooperatively with EEG during the same motor imagery period

the different movements of a same limb is recognized using decision fusion on a finer scale. The results of experiments reveal that the average recognization accuracy of three kinds of movement imagery of the right arm can be promoted from 63. 0% (only with EEG) to 91. 4% (with the proposed method). The proposed method may be applied in clinical in the future.