Development and Hybrid Control of an Upper Limb Prosthesis for Reach and Grasp Motions

Replacing the human upper limb with artificial devices of equal capability and effectiveness is a long-standing challenge. In this paper, a hybrid reach-to-grasp task planning scheme is proposed for transhumeral amputees exploiting both electromyography (EMG) and visual signals to control the upper limb prosthesis. EMG signals extracted from the subject are fed into the long short-term memery neural network to control the motion of the prosthesis after training and classification. The visual servoing module intends to detect and locate the object thus estimate grasping pattern in real time. In our control strategy, amputees are able to use the EMG signals to operate the prosthesis, and they can also activate the visual module at any moment, which recognizes and locates the object to be grabbed, and then moves the prosthesis close to the object and imposes grasping according the preset inference library, which reduces the cognitive and operational burden of amputees greatly. Finally, experiments are conducted on a patient with transhumeral left arm amputation to verify the effectiveness of the proposed control strategy using a upper limb prosthesis. The results showed that the hybrid control scheme brings more choices to control the prosthesis freely and flexibly.