Robust eye openness and head pose estimation based on multi modal camera

The paper proposes an approach to estimate the eye openness degree and 6DOF head pose with RGBD camera. A novel face localization method based on combination of 3D points cloud and RGB texture is proposed, which is robust against the interferences from background and foreground. The eye openness degree is estimated by a statistical model based on Mahalanobis distance, which is reliable even with glasses. The experiment results under different head poses are presented to evaluate the system's performance.