Google Research: Towards passive heart health monitoring via smartphone camera

7 hours ago

Google Research introduces a passive heart rate monitoring system (PHRM) that uses facial video from smartphone front cameras to measure heart rate (HR) and resting heart rate (RHR) during everyday use.
PHRM achieves mean absolute percentage error (MAPE) < 10% for HR across all skin tones and mean absolute error (MAE) < 5 bpm for RHR compared to wearables, setting new standards for remote photoplethysmography (rPPG).
The system was developed using a diverse dataset of over 350,000 video clips from nearly 700 participants, ensuring inclusive performance across skin tones via the Monk Skin Tone scale and non-inferiority criteria.
In free-living studies, PHRM outperformed 15 leading rPPG models, with MAPEs of 5.04% (light skin), 5.12% (medium skin), and 7.84% (dark skin), and successfully estimated RHR on 73.6% of participant-days.
PHRM-derived RHR correlates with cardiovascular risk factors like high BMI and low VO2max, validating its clinical relevance for passive health tracking.
Future improvements could address lower measurement success rates in darker skin tones and mitigate errors from motion or talking via camera optimization and video stabilization.
Google is releasing the dataset and a pre-trained model (PHRM-mini) to qualified researchers for non-commercial use to advance inclusive health monitoring research.

Hasty Briefsbeta