3D Facial Gaze Tracking Method using Kinect

摘要

In this paper, we propose a new facial gaze tracking method for large-size displays. The proposed method can track facial gaze accurately and promptly compared with previous methods. In an image captured by a Kinect v2, the face key point is detected by using the Kinect v2 High Definition face model. The plane that passes through the key point is calculated, along with the normal vector to the plane. Last, the facial vector parallel to the normal vector and passing the center of the plane is calculated. The intersection point between the facial vector and display is adopted for performing facial gaze tracking.