Hand and finger control of myo-prosthesis based on motion discriminator and voluntary control

In this study, we aim to control flexion and extension movements of thumb and fingers of a myoelectric prosthesis with myoelectric signal. In this research, it is strived to control extension and flexion movements of 4 fingers and thumb of the myoelectric prosthesis by a hybrid method, which conbine movements identification and joint angle estimation. Support Vector Machine, SVM, is used to identify movements, and Multi-Input and Multi-Output Nonlinear ARX model, MISO-NARX model, is used to estimate joint angle. As a result of the identification, identification rates of 5 movements have been more than 80 %. We have been estimated joint angles of thumb, index and pinkie in movement 1. As a result of estimating, the error of mean square of each fingers have been 0.0250144 (thumb), 0.251264 (index) and 0.403289 (pinkie). According to the results, the hybrid method is enabled to estimate finger joint angles. In the future, we will verify that the hybrid method is enabled to control myoelectric prosthesis.