Simultaneous Multi-Joint Myoelectric Control of Transradial Prostheses

By Christopher L. Pulliam As the development of dexterous prosthetic hand and wrist units continues, there is a need for command interfaces that will enable a user to operate these multijoint devices in a natural, coordinated manner. Commercially available myoelectric control algorithms are unintuitive, allow only a single degree of freedom to be operated at a time, and typically require some type of a mode switch to transition between operating the various functions of the system. Pattern recognition-based control algorithms take advantage of distinct patterns of activity related to the intuitive movements of the missing limb, thus reducing the cognitive burden associated with traditional control methods. To date, however, the majority of these advanced approaches still limit users to sequential operation of the various prosthesis functions. This project aimed to develop an intuitive myoelectric command interface that would allow multiple joints (forearm pronation-supination, wrist flexion-extension, and hand opening-closing) of a transradial prosthesis to be controlled in a highly coordinated fashion. We demonstrated than a time-delayed artificial neural network controller could be trained to accurately predict motions of several key hand and wrist motions using myoelectric signals recorded from muscles in the forearm offline. The controller was then evaluated in real-time during simulations of several tasks in a virtual