Abstract: A 5-DOF exoskeletal rehabilitation arm for hemiplegic upper limbs is developed with passive and active interactive exercise control strategy. In passive exercise, surface electromyogram (sEMG) of hemiplegic patients’ healthy limbs are extracted to control the arm because they are impaired unilaterally in general. AR model and BP neural network are used to understand the patient’s motion intention in order to actuate the arm. Then, it can assist disabled arm to implement preprogrammed motions. In active exercise, the torque of each joint is used to estimate the force caused by limbs’ movement in real-time. Terminal velocity is controlled with proportional controller and Jacobian inverse matrix to drive each joint. Test results prove the correctness and effectiveness of the proposed method.

Key words: AR model, BP neural network, Force assistant control, Rehabilitation arm, Rehabilitation strategy, Surface electromyogram