Homeokinetic prosthetic control: collaborative selection of myosignal features

We present an approach to the control of myoelectric prostheses that is based on a collaborative interaction of a prosthesis and the patient. During training, the patient is instructed to generate control signals that follow the observed autonomous movements of the prosthesis. At the same time, the prosthetic controller maximizes both the diversity of movements and the coincidences of prosthetic movements and human control signals by varying the features and control actions. This dual control principle which is derived from the homeokinetic robot control is demonstrated to be efficient for the control of a hand prostheses with two degrees of freedom.