This study designed an intelligent medical rehabilitation training system based on motion coordination, which is mainly aimed at patients with balance functions and motion coordination disorders. The system consists of training beds, pedal components, manual control devices, main control computers and corresponding training software. Biofeedback control through the brain is combining physical and psychological treatment of patients, artificial intelligence judgment and adjustment, analysis and evaluation of biofeedback. The rehabilitation device adopts the structure mode of the upper and lower computers. The lower computers take DSP as the core and cooperate to complete the functions of weak EMG signal detection and feedback, stimulation signal output and data transmission. The weak electrical signals remaining in the limbs of stroke patients are processed by the host to form a closed-loop biological feedback electrical stimulation system. According to the intermittent distraction joint method and aerobic training method, an intelligent medical rehabilitation device is designed, including a rehabilitation device for training the rotation function of the wrist and a rehabilitation device for training the flexion and extension function of the wrist. In order to meet the rehabilitation needs of the spasm and recovery periods, both wrist rehabilitation devices have three rehabilitation modes, namely passive rehabilitation mode, assisted rehabilitation mode and damping rehabilitation mode. The intelligent medical rehabilitation training instrument system can realize the combination of patient adjustment force and hand control training to achieve the matching of lower limb, lower limb movements and visual information, so as to promote the patient's balance function and coordinate rehabilitation training. The system sets visual and auditory feedback, and increases the interest of training by adjusting the training speed and training difficulty, which has important clinical application value.