IDENTIFICATION OF FINGER OPERATION USING SUPPORT VECTOR MACHINE AND CONTROL OF MYOELECTRIC PROSTHETIC HAND BASED ON INTEGRATED ELECTROMYOGRAM
Abstract Myoelectric prosthetic hands, which use surface electromyograms (SEMG) to identify the intended motion and the control movement of the artificial hand accordingly, have been studied for many years. Various signal processing and identification methods have greatly expanded the possibilities for studying myoelectric hands and many recent studies have been practical, with a focus on commercialization. In this study, the identification of finger operation and control of the fingers of a myoelectric prosthetic hand are discussed. An integrated electromyogram (IEMG) is used to extract features from finger operations, and a support vector machine-based classifier (SVM) is employed to identify the six different types of finger operation. To enhance the myoelectric prosthetic hand operability, two new control methods are proposed. In the first method, the prosthetic finger angle is changed based on the duration of the muscular activity during finger operation. The strength of the muscular activity is not reflected. Thus, the target angle of the finger is determined without considering the degree of the muscular activity. In the second control method, the prosthetic finger angle is changed based on not only the duration but also the degree of the muscular activity of the finger during finger operation. A contribution of this paper is that the formula to give a target angle of the finger in which the identification result of finger operation, the duration of the muscular activity and the degree of the muscular activity are reflected, was established. As each finger is controlled based on the degree of muscular activity, the second method enables users to adjust the prosthetic hand’s grasp on objects using their intent. Therefore, this method facilitates the grasping of complex-shaped objects. Although some commercial myoelectric prosthetic hands can adjust finger speed in proportion to the strength of muscular activity, in these systems all the fingers are controlled collectively. The proposed method has the advantage of enabling the fingers to be controlled independently on the basis of the strength of the muscular activity of each finger. This is achieved by incorporating the results of the finger operation identification into the target angle of the prosthetic finger. Both of the proposed methods for the myoelectric prosthetic finger control have been evaluated experimentally.