ラバン理論に基づいたヒューマンフォームロボット身体動作の動作特徴抽出と表出感情推定

Abstract

A set of movement feature values, called a Laban's feature value set, is proposed here in order to explain observers' impression of bodily expression. The design concept of the Laban's feature value set is based on Laban Movement Analysis (LMA), which is a proven theory in body movement psychology. In this paper, we adopt a Human Form Robot (HFR) as an agent of bodily expression because this is the type of a robot better adapted to Human-Agent Interaction (HAI). Relations between Laban's feature values and HFR's emotions (Happy, Angry, Sad and Relaxation) which are the subjects of this study were examined using the analysis of correlations. By the consideration of Russell's circumplex model of affect, we discussed the correlation in terms of each axial (“pleasure-displeasure” and “degree of arousal”) characteristics. Next, principal component analysis was conducted to examine multi-dimensional correlations. Finally four estimated emotion equations are generated by using the Laban's features.