Development of a robust intention estimation algorithm and rehabilitation technique based on muscle activation for prosthetic hand control

Abstract

To boost the usefulness of the prosthesis for upper limb amputation patients, it is necessary to enable independent and intuitive control of each finger degree of freedom. The control method of a commercial prosthetic arm has an advantage in stably performing a motion determined based on pattern recognition, but intuitive control is not possible. To overcome this, a study on the relationship between the model-based surface EMG and finger force intentions was conducted to enable simultaneous and proportional control of the fingers. High finger force intention estimation accuracy and independence were secured, but rapid performance decline was caused by changes in model generation and conditions such as release and wearing of prosthetic arm, change of electrode position, change of muscle condition, etc. This is a limitation of the principle of creating a training-based model, and in this dissertation, a neurophysiological model of surface EMG generation was created through parameter definition and formula to overcome the above limitation. Through the modified model, the activity of each brachial muscle is estimated from the surface EMG and used as intention estimation information. In addition, the robustness of the model was secured by designing a parameter adaptation algorithm according to changes in conditions. Another approach is to systematize the prosthetic adaptation process of patients with upper limb amputation by proposing a rehabilitation method through visualization of the estimated muscle activity. The performance of the proposed method was verified by comparing it with the preceding representative model, and the usefulness of the method was verified through real-time electric prosthetic control.