How Can Your Face Reveal Heart Rate? Exploring rPPG Technology

Background

By recording a short facial video, it is possible to measure heart rate and even respiration rate using remote photoplethysmography (rPPG), which captures subtle skin color changes caused by cardiac pulsation. This technique has important applications in health monitoring.

Algorithm Principle

The measurement works because the dermis and subcutaneous layers of facial skin contain abundant capillaries. When the heart beats, blood volume in these capillaries changes periodically, causing periodic variations in light absorption and thus skin color. Although invisible to the naked eye, a camera can capture these variations.

Measurable Physiological Signals

rPPG can recover the blood volume pulse (BVP) signal from facial video. After extracting raw color changes, the ratio of DC to AC components of different color channels can estimate blood oxygen saturation (SpO₂). Frequency‑domain analysis of the BVP yields heart rate and respiration rate, while peak detection provides inter‑beat intervals (IBI) for instantaneous heart rate, respiratory rate, and heart‑rate variability.

Existing Methods Overview

Methods are divided into traditional hand‑crafted approaches and data‑driven approaches. Hand‑crafted methods model skin reflectance or physiological signal physics to extract periodic signals, offering controllability in constrained scenarios but limited robustness to motion and lighting changes. Data‑driven methods learn mappings from color‑change features to physiological parameters using machine learning or deep learning, achieving higher performance in complex scenes but requiring large, diverse training data.

Algorithm Design

To ensure controllability, a traditional hand‑crafted pipeline is adopted. After obtaining raw R/G/B signals, an optimal color‑space projection combines channels to suppress motion artifacts. A quality‑based signal combination and algorithm restart strategy are added to adapt to various scenarios. The resulting algorithm is demonstrated in the “Qixi Heartbeat Test” effect.

References

Hao‑Yu Wu, Michael Rubinstein, Eugene Shih, John Guttag, Fredo Durand, William T. Freeman. “Eulerian Video Magnification for Revealing Subtle Changes in the World.” ACM Transactions on Graphics, 2012.

牛雪松, 韩琥, 山世光. “基于rPPG的生理指标测量方法综述.” 中国图像图形学报, 2020.