EMG Classification by using Swarm Intelligence for Myoelectric Prosthetic Hand

Abstract

In recent years, myoelectric prosthetic hand (MPH) has been extensively studied due to the spread of 3D printers. However, it cannot do precisely movement now because it is difficult to identify electromyogram (EMG) by using existing method. The reasons for this are as follows; Hand movement is too complicated to use it as label for supervised learning method, EMG change its characteristics gently with time. Accordingly, we need to develop a new method adapted to MPH. In this study we developed an identification method using swarm intelligence which was optimized to the characteristic of EMG. To verify the function of the method, experiments were conducted. For some subjects, identification rates were high. Moreover, we discussed how to improve the method and conducted some experiments to verify it. It has been considered effective to investigate the optimization method of particle swarms.