Abstract: For the rehabilitation training of hemiplegia patients caused by stroke,an upper-limb rehabilitation robot system based on motor imagery electroencephalography(EEG)is designed.Firstly,three-dimensional animation is used to stimulate the patient to imagine the upper-limb movement,and EEG signal is acquired by EEG amplifier via universal serial bus(USB). Secondly,eigenvector is extracted by wavelet packet transform(WPT),and linear discriminant analysis(LDA)classifier based on the Mahalanobis distance is utilized to classify the pattern.Finally,PC gives the visual feedback information based on virtual reality and controls the rehabilitation robot.The patient's upper-limb motor imagery EEG is used to control rehabilitation robot directly and it can accelerate the recovery of motor nerve function.Six subjects have been tested for a long time using this system.The results show the feasibility of the whole system.