Unconscious gestures that empower turn taking for telepresence robot

Abstract

In this research, we propose a telepresence robot that avoids speech collisions occurs in remote conversations. In face-to-face conversations, humans predict the next speaker by seeing others' gestures. However, it is difficult to predict it at 2D video chat situations. The failure of prediction causes speech collisions and awkward conversations. To solve the problem, we propose the telepresence robot that conveys 3D spontaneous gestures. The proposed system employs Kinect as a gesture input device and conveys unconscious gestures to telepresence robot. We conducted experiments with subjects and confirmed the efficiency of the proposed system.