Telepresence robot conveying pre-motions for avoiding speech collisions in teleconference

In this paper, we propose a telepresence robot for avoiding speech collisions occurred in remote conversa- tions. In face-to-face conversations, humans predict who will speak in next by referring other participants´ motions. These motions are called pre-motions. However, pre-motions are likely to be unperceived in teleconferences such as video chat conversations using two-dimensional video. The failure of the prediction causes speech collisions. It undermines the participants´ motivation and wastes time. To solve this problem, we propose a telepresence robot conveying gestures including pre-motions. We use Kinect for the gesture perception device and humanoid robot for conveying unconscious gestures.