EMG-Based Interface Multi-degree of Freedom and Optionality

Abstract

We aim to develop a myoelectric interface, which is excellent at voluntary and multi-degree of freedom. In this paper, joint angle of a wrist is estimated by myoelectric signal for the purpose. In previous study, NARX model has been a method applicable to estimate joint angle of a wrist. Multi-input and multi-output NARX model, MIMO-NARX model, is selected to develop a interface because it is able to estimate angle of multi-degree of freedom. Moreover, Bayesian network is used to decide orders of MIMO-NARX model. As a result, MIMO-NARX model has been created to estimate volar flexion- dorsal flexion motion and pronation-supination motion of forearm, and then it has been estimated with the error of mean square as 10-1 orders. The Bayesian network had been defined a causal relationship between myoelectric signal and joint angle of a wrist and it has been clarified the impact of Electric-Myo-Delay, EMD, on dynamic characteristics. In the future, estimate accuracy will be improved by entertaining the domain decomposition method again because the estimate accuracy is not enough. Moreover, NARX model will be developed by considering EMD and defining a causal relationship between myoelectric signal and hand joint.