Optimal Mapping of Torus Self-Organizing Map for Human Forearm Motions Discrimination on the Basis of Myoelectric Signals

Abstract

This paper describes an optimal mapping of the torus self-organizing map for a human forearm motion discrimination on the basis of the myoelectric signals. This study uses the torus self-organizing map (Torus-SOM) for the motion discrimination. The normal SOM identify input data into the same feature group by using the all units of map. Then there is a possibility of the misrecognition motion around the boundary lines of the motion groups. Therefore, this study proposes the mapping method of SOM that the learning units of the same motion concentrate on one local range and the learning unit groups of each motion separates enough. As a result, the variance in the same motion group becomes small and the variance between each motion groups becomes big. Some experiments on the myoelectric hand simulator show the effectiveness of the proposed motion discrimination method.