Development of a Attentive Listening Robot Using the Motion Prediction Based on Surrogate Data

The development of nonverbal behavior functions such as nodding is also important for natural dialogue robots. However, since natural nonverbal dialogue functions is not developed, it is not possible to gather natural human-robot dialogue data. In this study, we developed the attentive listening robot system that uses human-human dialogue data as surrogate data. The nodding prediction model substitute human-human dialogue data for human-robot dialogue data to predict the behavior of the dialogue robot. The proposed system makes a judgment on whether to nod based on the output of the prediction model, audio and image information are input to the model. The results of the attentive listening experiment suggested that the proposed system that generates noddings that make it easier to speak was developed.