An auxiliary gaze point estimation method based on facial normal

Area of Science:

• Computer Vision
• Human-Computer Interaction
• Biomedical Engineering

Background:

• Existing gaze point estimation methods often restrict head movement, potentially causing discomfort or injury.
• There is a need for gaze tracking systems that are accurate, robust, and non-intrusive.

Purpose of the Study:

• To propose a novel gaze point estimation method utilizing facial normal and binocular vision.
• To overcome the limitations of existing methods by allowing unrestricted head movement.

Main Methods:

• Stereo camera calibration to obtain intrinsic and extrinsic parameters.
• Face, iris, nostril, and mouth detection using Viola-Jones and integro-differential operators.
• 3D coordinate calculation via stereo matching and reconstruction.
• Facial plane approximation using least squares fitting.
• Gaze point determination by intersecting the facial normal with the computer screen.

Main Results:

• The proposed method accurately estimates the point-of-gaze.
• Experimental results demonstrate the robustness of the technique.
• The system allows for natural, unrestricted user head movement.

Conclusions:

• The developed gaze point estimation method is accurate and robust.
• This approach offers a more comfortable and less injurious alternative to current gaze tracking systems.
• The method shows promise for applications requiring natural interaction.