In this paper, we present a new system to track eye gaze using a single camera and single light source. In system implementation, an accurate and robust eye gaze tracking eigenvalue extraction algorithm based on image analysis was proposed. The system successfully positions glint center and pupil center at sub-pixel level, and it remains a relatively high accuracy and speed in different illumination situations. Experiments show the average variance of the system is approximately 21 pixels. The method has increased the accuracy level of single-camera eye gaze tracking system, and offered a both efficient and low-cost solution for eye -gaze tracking oriented human-computer interaction.