An assistance training control system based on the prototype of an upper limb rehabilitation robot with three degrees of freedom was presented.The system was designed to detect the patients' real-time intention of moving and realize the assistance training.A variety of electronic technologies were used to capture the surface electromyogram (sEMG) signal of radiocarpus and cubitalis anterior.The sEMG signal was captured and processed by a multi-core hierarchical motion control system.Then, the assistance training was conducted according to the moving intention of the patient.Finally, the sEMG signal identification experiments and speed regulation experiments were performed and the feasibility of the assistance training control system was verified.