Emotion Estimation from Robot Movements Using Weighted Random Forest and Pairwise Classification

Recently, various communication robots are developed. These robots have functions to express emotions. Emotions are influenced by verbal and nonverbal information. Nonverbal information including facial expression and body movements has a greater influence on emotional interaction than verbal information has. In this paper, we focus on body movement, which is one of way for a robot to express its emotions for communication using nonverbal information. Emotions are expressed with some basic emotions. We defined two emotions, a primary emotion and a secondary emotion. We propose an estimation method using weighted random forest and pairwise classification. We confirm effectiveness of the proposed method.