Hierarchical Human Motion Recognition By Using Motion Capture System

Service robots are requested to move promptly according to the motion of human. For that, it is necessary to realize accurate recognition of motion. This paper proposes a method to hierarchically perform rough motion recognition and fine movement recognition. First, various kinds of actions are roughly classified by threshold processing based on a rule base. After that, use SVM only for limited information and classify it finely. By this method, it can be expected that SVM can be effectively used to recognize the operation. In order to confirm the performance of the proposed method, we classified the motion data using the proposed method. As a result, the proposed method was able to discriminate the operation with an accuracy of 77.6%. The present result suggested that it becomes possible to apply to a human support robot by increasing the corresponding action.