Avatar

We examined the effect of incorporating gaze aversion behavior during turn-taking in dialogue agents, demonstrating that it creates intelligent and natural impressions. The importance of nonverbal gestures was clarified through mock interviews and dialogue experiments.

In a VR virtual audience system using a speech-driven nodding generation model, we found that applying the nodding frequency adjusted for a single virtual audience member to multiple audience members results in excessive nodding. Through experiments, we demonstrated that reducing the nodding frequency as the audience size increases creates a more natural impression.

When operating 3DCG avatars with VR equipment, it is difficult to input poses suitable for avatars with different appearances from real-world users due to user skill and device constraints. In this research, we proposed a support system that allows users to input target poses easily and accurately by continuously blending between target poses and user input postures.

FlexFace is a head gesture display system for telecommunication. Traditional video communication systems use rigid flat screens, which feel inorganic. FlexFace applies the principles of "squash and stretch" from animation techniques to create more lifelike movements by deforming a flexible screen.