A Study on Human Gaze Detection Based on 3D Eye Model

摘要

Human gaze can give valuable tips for human computer interaction, but it is very difficult to detect human gaze position with one or two camera systems. Conventional method has the limitation of inaccurate gaze detection performance or not being able to track the fast motion of user's face and eye. To overcome such problem, in this paper, we propose a new gaze detection method. Compared to previous works, our method has following three advantages. First, our method uses three camera systems, such as a wide and narrow view stereo cameras and allows user's natural head and eye movement. Second, to obtain gaze position on a monitor, we detect the 3D eye position and gaze vector of eyeball. Third, to enhance the eye detection performance, we use AdaBoost eye detector and PCA algorithm. Experimental results showed that our method could be used for real-time gaze detection system.