We implemented a gaze algorithm for interacting with multiple observers as a precursor to a multi-party conversation system. By acknowledging multiple participants in a natural manner, we seek to set the stage for smo...
详细信息
ISBN:
(纸本)9783319474373;9783319474366
We implemented a gaze algorithm for interacting with multiple observers as a precursor to a multi-party conversation system. By acknowledging multiple participants in a natural manner, we seek to set the stage for smoother and more effective human-robot conversations featuring proper turn-taking using attention shifts. The android robot EveR-4, developed at the Korea Institute of Industrial Technology for human-robot interaction (HRI) applications was used. The robot wore a dress and was made up to replicate interacting naturally with a real woman as much as possible. Using a RGB-D camera, peoples faces and positions were tracked so that the robot's attention could be given to everyone appropriately. An importance value was assigned to each detected face based on the length of time it was detected and its distance to the robot. Facial expressions were made by the robot when people were seen to increase observers' sense of interaction. We observed peoples reactions to our implementation at an exhibition and made note of how we can improve the overall system to be more life-like and realistic.
暂无评论