Motion Analysis in Vocalized Surprise Expressions and Motion Generation in Android Robots

Surprise expressions often occur in dialogue interactions, and they are often accompanied by verbal interjectional utterances. We are dealing with the challenge of generating natural human-like motions during speech in android robots that have a highly human-like appearance. In this study, we focus on the analysis and motion generation of vocalized surprise expression. We first analyze facial, head, and body motions during vocalized surprise appearing in human-human dialogue interactions. Analysis results indicate differences in the motion types for different types of surprise expression as well as different degrees of surprise expression. Consequently, we propose motion-generation methods based on the analysis results and evaluate the different modalities (eyebrows/eyelids, head, and body torso) and different motion control levels for the proposed method. This work is carried out through subjective experiments. Evaluation results indicate the importance of each modality in the perception of surprise degree, naturalness, and the spontaneous versus intentional expression of surprise.