RESUMO
This paper introduced a hand function rehabilitation system based on motor imagery (MI) brain-computer interface for hand function rehabilitation of stroke patients. The rehabilitation system contains three subsystems. Offline training subsystem displays the blank screen, a left or right hand movement video and arrow in turn, which respectively reminders patients to rest and make preparations for MI and instruct them how to do MI, and be doing MI. Finally, the patients' electroephalography (EEG) signals are acquired and processed togenerate a recognition model. Model update online training subsystem presents the black screen and a left or right arrow, the meanings ofwhich are the same as those in offline training subsystem. Then the acquired EEG signals are analyzed according to the established recognitionmodel. Next, the analysis result is derived to control the hand movement video to be played. The video can also act as a visual feedback,which makes patients' EEG signals easier to be recognized. The updated and more effective recognition model is built at last. Virtual reality(VR) online training subsystem constructs 3D grid models of VR scene, a 3D man model and its hand animations in the 3Dmax. Then, all ofthem are imported into Unity3D. The control methods of the animations are also designed in Unity3D. In the end, the patients' EEG signalsare analyzed according to the updated recognition model, thus controlling the hand movements of the 3D man in real time. The developedsystem has many characteristics, such as multilevel training and more immersion, which hopefully promotes the plasticity of central nervoussystem. The designed system provides new treatments for post-stroke hand function rehabilitation and further lays the foundation for family-mode rehabilitation.