Balancing cardiac privacy with quality in video recordings

Area of Science:

• Biomedical Engineering
• Computer Vision
• Signal Processing

Background:

• Remote photoplethysmography (rPPG) extracts heart rate from facial videos.
• rPPG technology raises privacy concerns due to potential misuse of sensitive physiological data.

Purpose of the Study:

• To develop a reversible video modification framework for secure rPPG signal transmission.
• To enhance privacy by concealing heart rate information while preserving video quality.

Main Methods:

• A novel framework using frame-wise sinusoidal modulation on facial regions to remove, encrypt, transmit, and restore rPPG signals.
• Evaluation using seven rPPG techniques on the LGI-PPGI dataset with metrics like PSNR, SSIM, and correlation.
• Comparison against prior methods for signal suppression and visual fidelity.

Main Results:

• The proposed method achieved an overall score over 0.75, outperforming previous approaches by approximately 50%.
• High visual fidelity was maintained (PSNR ≈ 68, SSIM ≈ 0.97).
• Physiological information was effectively concealed, increasing heart rate estimation error by 22 bpm compared to prior methods.

Conclusions:

• The study presents the first end-to-end reversible framework for privacy-preserving video transmission of facial recordings.
• The lightweight and effective approach is suitable for real-time applications like video conferencing and telehealth.