In view of the existing problems in rehabilitation training in China and the great potential of rehabilitation training robots based on Motor Imagery Brain Computer Interface, this paper designs and develops a wearable knee rehabilitation training system based on Motor Imagery Brain Computer Interface from the perspective of portability and low cost. The proposed knee rehabilitation training system can extract the patient’s active rehabilitation intention through EEG signal, and complete the active and passive rehabilitation training of the knee joint through the knee exoskeleton, which can effectively overcome the problems of the traditional rehabilitation training system, such as high training cost, limited number of professional rehabilitation therapist, and low willingness of the patient’s active training. A series of offline and online experiments verify the feasibility and effectiveness of the proposed rehabilitation system. The development and application of this system is expected to provide a rehabilitation system with autonomous control, active rehabilitation and convenient wearing for patients with knee joint dysfunction, and effectively alleviate the problems existing in the current motor function rehabilitation training.