Arm prosthetic control through electromyographic recognition of leg gestures

Use of surface electromyography (sEMG) has shown promise as a method of controlling prosthetic devices, but current control paradigms may suffer from at least one of two main problems: non-intuitive input and invasiveness. A typical approach to the problem is to recognize gestures by recording EMG signals from multiple muscle sites near the amputation and analyzing these signals with a classification algorithm to detect a discrete user intention (e.g. close hand, rotate wrist, etc.). Here, we introduce the idea of recording EMG from the lower leg as the control input to a prosthetic arm. The user performs gestures with the lower leg or foot which map relatively intuitively to forearm or hand movements, requiring essentially no training. A completely noninvasive system for recording, processing, and transmitting commands from the leg to a prosthetic arm could be reduced to a sleeve design with embedded electronics. Although the musculature of the lower leg is somewhat different from that of the forearm on which most EMG-based gesture recognition research is performed, we show that many leg and foot gestures, which have directly analogous arm and hand movements, can be recognized with pattern recognition techniques.