EMG-Based Control for Three-Dimensional Upper Limb Movement Assistance Using a Cable-Based Upper Limb Rehabilitation Robot

Voluntary residual motor efforts to the affected limb of patients after stroke have not been involved enough in most rehabilitation robot control strategies. In this paper, a natural integration between human and machine is proposed by using the surface electromyography (EMG) signals from six muscles which mainly contribute to the upper limb movement. A linear state space model is trained, which can estimate the real-time movement intention by using EMG signals, to calculate the movement needed forces and then provided by a cable-based upper limb rehabilitation robot. Ten healthy subjects are recruited to complete the tasks with and without robot assistances. The performances of the subjects with the assistances are compared to that of the subjects without assistances. Results show that the forces from the model were real-time continuously estimated and accurate. Furthermore, there is no significant difference in the group mean root mean square error (RMSE) and muscle activations between the task without assistance and with assistance. These results show that the robot using the state space model could provide physiologically appreciate assistance to the subject, and the robot could conduct the rehabilitation training combined with the voluntary residual motor efforts. Clinical test will be carried out to validate the feasibility of the robot-aided rehabilitation using myoelectrical control.