Robust heart-rate estimation from facial videos using Project_ICA

Area of Science:

• Biomedical Engineering
• Signal Processing
• Physiological Monitoring

Background:

• Remote photoplethysmography (rPPG) enables non-contact heart rate (HR) measurement from videos.
• Current rPPG methods are limited by facial motion and head movements.

Purpose of the Study:

• To develop a robust system for remote HR estimation from facial videos under various movement conditions.
• To address limitations of existing rPPG techniques.

Main Methods:

• Developed a system framework for remote HR estimation.
• Employed 'Project_ICA', a method based on a skin reflection model, to extract pulse signals.
• Created a dataset of 112 videos with diverse participants and conditions.

Main Results:

• The Project_ICA method demonstrated more accurate and robust HR estimates compared to existing methods.
• Performance was evaluated across various skin tones, body motions, and post-exercise HR recovery.
• Challenges remain in obtaining reliable measurements from individuals with darker skin tones.

Conclusions:

• The proposed Project_ICA method offers a significant advancement in non-contact HR monitoring.
• Further research is needed to overcome challenges related to diverse skin tones in rPPG.