Upper Limb Prosthesis Control for High-Level Amputees via Myoelectric Recognition of Leg Gestures

Recognition of motion intent via surface electromyography (EMG) has become increasingly practical for prosthesis control, but lacking residual muscle sites remains a major obstacle to its use by high-level amputees. Currently, there are few approaches to upper limb prosthesis control for individuals with amputations proximal to the elbow, all of which suffer from one or more of three primary problems: invasiveness, the need for intensive training, and lacking functionality. Using surface EMG sensors placed on the lower leg and a natural mapping between degrees of freedom of the leg and the arm, we tested a noninvasive control approach by which high-level amputees could control prosthetic elbow, wrist, and hand movements with minimal training. In this paper, we used able-bodied subjects to facilitate a direct comparison between control using intact arm and leg muscles. First, we found that foot gestures could be classified offline using time domain features and linear discriminant analysis with accuracy comparable to an equivalent system for recognizing arm movements. Second, we used the target achievement control test to evaluate real-time control performance in three and four degrees of freedom. After approximately 20 min of training, subjects tended to perform the task as well with the leg as with intact arm muscles, and performance overall was comparable to other control methods.