Gaze Detection Towards a Vision Switch
Yasuyuki YAMAZAKI, Hiroshi Mizoguchi, Kazuhiro Fukui · IEICE technical report. Speech · 2013
In this paper, we propose a method for detecting visual attention towards a ”Vision Switch”, which is a new concept of intelligent switch. It can be realized as an intelligent switch to control various devises without contact. In conventional method, visual attention is detected based on the result of gaze estimation. On the other hand, we handle gaze detection as a 2-class classification problem of attention and non-attention. Since it is difficult to detect visual attention with only an image, we apply statistical pattern recognition to movies. In addition, to detect user’s visual attention accurately, we should consider both the head and eye direction. Thus, we connect images of around eyes and whole face, then classify the connected images. The effectiveness of the proposed method for detecting visual attention is demonstrated through several basic experiments.