In recent years, the number of opportunities to see a people on video call through an agent or a monitor displayed on a display has increased. Along with that, the number of in-screen guidance agents that interact with people is increasing. In a one-to-one interaction with an in-screen agent, the listener and the speaker are fixed, so there is no need to take any other person into account. However, in the case of a conversation between two or more people such as in telepresence robot, it is problematic to speak to anyone in the vicinity. It is difficult for an agent in the screen or a person in the monitor to look out of the monitor due to the Mona Lisa effect. There is a lot of research being done to turn people on the screen to look. In this research, we focus on the connection of communication channels and propose an in-screen agent that talks with a specific person considering that is engaging in a multi-person conversation. We proposed and evaluated a method to follow the gaze target with the background of the agent’s monitor and a method to give inertia to the body posture when following. We also proposed and evaluated a method of expressing the gazing target by directly reflecting the gazing target on the pupil. In the study of the in-screen agent that can perform gaze, the gaze target is evaluated, but it does not verify what the conversation participants other than the gaze target think. In this study, the method of projecting the gaze target on the pupil and the method of the previous research are developed, and a comparative experiment is performed.
Proposal of Communication Channel Establishment Method with In-Screen Agent in Multi-Person Dialogue：fly
