Alphabet Validates Smartphone Heart-Rate System on 162,546 Videos Across 3 Skin-Tone Groups
Updated
Updated · Nature.com · Jun 1
Alphabet Validates Smartphone Heart-Rate System on 162,546 Videos Across 3 Skin-Tone Groups
1 articles · Updated · Nature.com · Jun 1
162,546 validation videos from 211 participants showed Alphabet’s passive smartphone system could measure heart rate during normal phone use with 6.09% participant-level MAPE, below the 10% industry benchmark.
Three skin-tone groups all met that accuracy target, and the darkest group remained within a prespecified non-inferiority gap, addressing a longstanding weakness of camera-based photoplethysmography.
504 of 685 participant-days produced valid daily resting-heart-rate estimates, with 3.62 bpm day-level error versus a wearable tracker; from day 3 onward, accuracy in the darkest group also fell below 5 bpm.
192,353 development videos from 485 participants trained the model, which captures 8-second front-camera clips after screen unlocks and filters low-confidence readings before aggregating them into daily resting rates.
Alphabet said the results suggest smartphones could broaden access to longitudinal heart-health tracking beyond wearables, and it released a pretrained model and annotated dataset for further research.
Your phone may soon watch your heart. Where is the line between life-saving tech and perpetual surveillance?
As Alphabet's heart-rate AI seeks billing codes, could your smartphone data soon affect your health insurance rates?
Alphabet's AI promises to close the health gap, but can it overcome the global digital and infrastructure divide?
Alphabet’s AI-Powered Passive Heart Rate Monitoring: Validation, Equity, Privacy, and the Future of Mobile Health
Overview
Alphabet, through Google and Fitbit, has introduced a breakthrough AI-driven system that passively tracks heart rate using ordinary smartphones. This deep-learning technology works in the background as users go about their daily routines, requiring no active input. The system was rigorously validated in June 2026, confirming its ability to deliver reliable heart rate data through an accessible device. This innovation marks a significant advancement over previous technologies, seamlessly integrating health monitoring into everyday life and highlighting the potential for continuous, user-friendly health insights.