Collection and analysis of multimodal interaction in direction-giving dialogues: Towards an automatic gesture selection mechanism for metaverse avatars

Abstract

With the aim of building a spatial gesture generation mechanism in Metaverse avatars, we report on an empirical study for multimodal direction-giving dialogues and propose a prototype system for gesture generation. First, we conducted an experiment in which a direction receiver asked for directions to some place on a university campus, and the direction giver gave directions. Then, using a machine learning technique, we annotated the direction giver's right-hand gestures automatically and analyzed the distribution of the direction of the gestures. As a result, we proposed four types of proxemics and found that the distribution of gesture directions differs with the type of proxemics between the conversational participants. Finally, we implement a gesture generation mechanism into a Metaverse application and demonstrate an example. © 2012 Springer-Verlag.