Remote Heart Rate Estimation Based on 3D Facial Landmarks.

Area of Science:

• Biomedical Engineering
• Computer Vision
• Physiological Monitoring

Background:

• Remote photoplethysmography (rPPG) enables non-contact heart rate (HR) estimation.
• Traditional rPPG methods often rely on 2D facial landmarks, limiting robustness to head pose variations and occlusions.
• Accurate and robust HR monitoring is crucial for telemedicine and health diagnostics.

Purpose of the Study:

• To develop a novel video-based remote heart rate estimation method utilizing 3D facial landmarks.
• To enhance the robustness and accuracy of HR estimation against head rotations and partial face occlusions.
• To introduce an effective face patch visibility check for improved signal quality.

Main Methods:

• Employing 3D facial landmark detection for video-based HR estimation.
• Implementing a novel face patch visibility check using 3D face patch normals.
• Experimentally comparing the proposed 3D method against 2D landmark-based baselines.

Main Results:

• The proposed 3D facial landmark method significantly improves HR estimation robustness to head rotations.
• The method demonstrates enhanced performance in the presence of partial face occlusions.
• The face patch visibility check effectively selects reliable facial regions, boosting HR estimation accuracy.

Conclusions:

• 3D facial landmarks offer superior robustness for video-based remote HR estimation compared to 2D landmarks.
• The proposed visibility check mechanism is critical for improving HR estimation accuracy in real-world scenarios.
• This novel approach advances non-contact physiological monitoring capabilities.