非接触計測による自己キャラクタ対面合成型実映像対話システムE-VChatの評価

摘要

Abstract – We previously proposed an embodied video communication system called“E-VChat” in which a CG avatar is superimposed on the other talker’s video image in aface-to-face scene in order to comprehend the talkers’ mutual interaction during remotecommunication. Remote talkers can communicate interactively via their own avatar withthe partner’s video image, which moves based on the talker’s motion and automaticallygenerates motion based on the on-off pattern of the talker’s voice. This system uses aheadset-type motion capture device that reflects the talker’s head movements directlyusing acceleration and gyro sensors. Further, we demonstrated the effectiveness of theE-VChat system through a communication experiment.　In this paper, we propose an advanced E-VChat system that uses image processingto sense the talker’s head motion without wearing sensors for practical use, and discussthe development of a prototype of this system. We demonstrate the effectiveness of theself-avatar that uses the talker’s head motion and auto-generated motion based on speechinput through a communication experiment in which 12 pairs of subjects are evaluatedin three separate communication modes. In one mode, only head motion is used; in thesecond mode, motion is auto-generated based on speech input; and in the third mode,both the head motion and the auto-generated motion based on speech input is used. Inaddition, we confirm through a communication experiment the effectiveness of the superimposedavatar for a face-to-face floating projection based on the partner’s face anglein comparison with a bottom-center projection of the partner’s image irrespective of thepartner’s angle.