Recognition of Objects Indicated by Deictic Pronouns for Helper Robots

Speech is a promising human-interface means for helper robots, which have growing needs in coming aging society. However, we cannot understand speech from verbal information alone in face-to-face communication. We may say "get that", even though we did not mention the object indicated by \´that\´ in previous talk. We can usually understand the object indicated by \´that\´ by vision in such cases. In our previous work, we proposed a speech interface using vision to understand such utterances. However, we did not pay much attention to demonstrative pronouns of deictic use, which are often used in our daily communication. Since pronouns are useful for understanding the position of target objects, we performed experiments to investigate how we use such pronouns. In this paper, we propose a method of recognizing objects indicated by deictic pronouns based on the experimental results