Abstract
Methods of EMG-based hand motion classification (EHMC) using classifiers have been developed to control multi-degree-of-freedom myoelectric hands. Though, for forearm amputees, effectiveness of the EHMC methods using classifiers have not been examined enough. This paper describes the effectiveness for seven forearm amputees who have no experience of training for the EHMC and have wide varieties of attributes. Their EMG signals were measured while they performed muscle contractions according to five kinds of motion examples presented in a predetermined order, and EHMC experiments were conducted with the EHMC methods using four different classifiers. As a result, all EHMC methods showed classification rate averaged across all subjects of more than 91%. In particular, the EHMC method using support vector machines showed the highest classification rate of 94.1%, and classification rate of each subject ranged from 98.0% to 90.9%. The results show that the EHMC methods have the effectiveness for amputees who have no experience of the training and have wide varieties of attributes.
