Clarification of fundamental motion using hierarchical clustering and graph theory

Recently, the development of motion-copying systems has enabled us to preserve and reproduce motion data including contact motion by humans. A system having multiple degrees of freedom (DOFs) is necessary for precise preservation of human motion. However, the DOF of a permeating device might not be the same as the DOF of the device used at the time of motion saving. The motion data should be reproduced precisely even if the system has a lower DOF than the system of the motion-saved phase. To address this problem, a method to determine redundant motion data on the basis of analysis of human motion is proposed. Hierarchical clustering is used for the analysis of motion data having high similarity. Furthermore, redundant motion data are determined for the maximum eigenvalue and eigenvector obtained from the adjacency matrix in graph theory. Using graph theory, the redundant data can be determined in each motion. The saved motion is reproduced by a lower-DOF system. The validity of the proposed method is confirmed by experiments in which grasping-motion data with four DOFs are reproduced by a 3-DOF device.