Similarity of human motion: congruity between perception and data

To duplicate human motion on the computer while retaining the individuality of the source, this paper discusses what the viewers focus on when they attempt to distinguish the identity of the source. In order to define the degree of similarity on motion, the data obtained by motion capture system are categorized using Self-Organizing maps, and the validity of the categorization is investigated to confirm the extent to which it is in agreement with the viewer´s intuitive judgment. Human walking patterns are measured and shown as animations of a skeleton model to permit distinction testing. The degree of similarity between walking patterns is calculated according to the results of the tests. The motion data are processed using Self-Organizing maps, and the degree of similarity from the aspect of data processing is also calculated as the distance between walking patterns in the map. These two measures of similarity are compared, and it is found that the degree of similarity between the human viewpoint and the data processing viewpoint indicates a high level of correlation.