Skip to Main Content
This paper presents requirements and a software framework for a humanoid robot aimed at speech interaction in a real world environment. We believe that for effective speech interaction with humans in the real world, not only is speech understanding and response generation required, but also other behavioral competencies are of paramount importance. Nodding, filler insertion, face tracking, and reactions to environmental stimuli during interaction are examples of such behaviors. In this paper, we discuss the importance of those competencies, termed” Naturalness Support Behavior”, and list the requirements for their implementation. A layered framework that satisfies the requirements is proposed. The system is integrated into the EGO architecture, which is the behavior control architecture for the Sony humanoid robot QRIO SDR-4XII. We also present preliminary experimental results of speech interaction based on the proposed ideas and conducted on QRIO.