Autonomous turn-taking agent system based on behavior model

Abstract

In this paper, we propose a turn-taking simulation system using animated agents. To develop our system, we analyzed eye-gaze and turn-taking behaviors of humans during actual conversations. The system, which can generate a wide variety of turn-taking patterns based on the analysis, will play an important role for modeling the behaviors at turn-takings, such as gazes, head orientations, facial expressions and gestures. The paper describes the system concept, its functions, and implementations. The findings obtained from investigations using the system will contribute to development of future conversational systems in which agents and robots communicate with users in a lively and emotional manner. © 2009 Springer Berlin Heidelberg.