Host: Japan Society for Fuzzy Theory and Intelligent Informatics (SOFT)
Human motions expressed in everyday life are composed of fundamental (primary) motions and modification (secondary) motions. While the primary motions includes e.g. walking, standing, sitting and any other operations by hands and legs, the secondary motions includes the motions features expressing exaggeration or suppression corresponding to human physical, physiological and psychological conditions, e.g. fatigue, sleepiness, confusion, pleasure, sadness, or dissatisfaction. In this study the motion data sequences obtained by 3D motion capture systems are transformed into trajectories in an eigenspace with time parameter, and then they are processed by HMM (Hidden Markov Models) to discriminate among ambiguous linguistic categories expressing the motion modalities. Some experiments using human subjects were conducted for investigating the effectiveness of the proposed method.
