A Novel Zernike Moment-Based Real-Time Head Pose and Gaze Estimation Framework for Accuracy-Sensitive Applications

Area of Science:

• Computer Vision
• Human-Computer Interaction
• Robotics

Background:

• Real-time head pose and gaze estimation (HPGE) is crucial for advanced human-machine and human-robot interactions.
• Applications like Driver's Assistance Systems (DAS) require accurate HPGE to prevent accidents.
• Existing methods often struggle with variations in illumination, background, and occlusion.

Purpose of the Study:

• To develop a novel hybrid framework for improved head pose and gaze estimation.
• To combine appearance-based and geometric-based methods for robust feature extraction.
• To enhance the accuracy and real-time performance of HPGE algorithms.

Main Methods:

• A hybrid framework integrating appearance and geometric-based approaches was proposed.
• Zernike moments were utilized for extracting rotation, scale, and illumination invariant features.
• Conventional discriminant algorithms were employed for classifying head poses and gaze direction.

Main Results:

• The proposed framework demonstrated accurate estimation across varying illumination conditions.
• Achieved an accuracy of approximately 85% for head pose and gaze estimation.
• Exhibited fast response times: 21.52 ms for head pose and 7.483 ms for gaze, independent of environmental factors.

Conclusions:

• The hybrid HPGE algorithm offers a promising solution for robust and accurate estimation.
• The method shows invariance to illumination, background, and occlusion, crucial for real-world applications.
• Future developments aim for enhanced robustness, including invariance to blurring conditions.