Key Points
We observe interactions between receptionists and users and incorporate awareness and response states into an image-based avatar. By replaying behaviors appropriate to each state at suitable times, the method aims to start interactions with users smoothly.
Abstract
An on-screen guide can be difficult to approach if it does not react when someone arrives or takes too long to respond. We investigated how an image-based avatar, made from video of a real person, can begin an interaction more naturally.
We first observed the behavior of receptionists and visitors at an information center. Using these observations, we gave the avatar two stages: noticing the user and responding to the user. The system starts an interaction automatically by replaying appropriate receptionist behaviors at suitable times in each stage.
In experiments, participants rated the beginning of interactions with and without these two stages. The version that included both stages received higher ratings. The results support designing not only the avatar's response, but also its behavior when it first becomes aware of a user.
Publication
- Tsubasa Miyauchi, Ayato Ono, Hiroki Yoshimura, Masashi Nishiyama, Yoshio Iwai,
Embedding the awareness state and response state in an image-based avatar to start natural user interaction,
IEICE Transactions on Information and Systems, Vol.E100-D, No.12, pp.3045 - 3049, December 2017.
