Google: Detecting Heart Rate Through Smartphone Camera AI

2026-06-17 · Source: AI Magazine · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Medical Devices & Health Technology, Emerging Technologies & Innovation · Depth: Advanced, short

Summary

Google researchers have developed a novel passive heart rate monitoring (PHRM) technology that uses smartphone front-facing cameras and deep learning to detect heart rate with medical-grade accuracy. This system analyzes facial video during normal smartphone use, achieving a mean absolute percentage error of less than 10%, meeting industry standards across all skin tones. Unlike previous methods requiring finger placement, PHRM operates in the background. The research involved the largest and most diverse remote photoplethysmography (rPPG) study to date, utilizing over 350,000 video clips from nearly 700 participants, with at least 33% representing dark skin tones. This approach addresses prior underrepresentation issues and outperforms 15 leading rPPG models, setting a new standard for daily resting heart rate estimation with wearable-level accuracy.

Key takeaway

For Machine Learning Engineers developing health monitoring applications, Google's passive heart rate monitoring research demonstrates a robust, equitable approach. You should prioritize diverse datasets, especially for underrepresented groups, to achieve medical-grade accuracy across all demographics. Consider integrating rPPG methods for background health tracking, leveraging on-device processing to enhance user experience and data privacy. This work sets a new benchmark for non-contact physiological measurement.

Key insights

Smartphone cameras can passively monitor heart rate with medical-grade accuracy across diverse skin tones using deep learning.

Principles

• Diverse datasets improve model robustness.
• Passive monitoring enhances user experience.
• rPPG can achieve wearable-level accuracy.

Method

An on-device software pipeline processes facial video clips, then temporal shift convolutional neural networks predict heart rate by detecting light interaction with skin.

In practice

• Integrate PHRM for background health tracking.
• Use diverse datasets for equitable AI models.
• Explore camera exposure optimization.

Topics

• Passive Heart Rate Monitoring
• Remote Photoplethysmography
• Deep Learning
• Smartphone Health
• Algorithmic Bias Mitigation
• Cardiovascular Biomarkers

Best for: Computer Vision Engineer, AI Product Manager, AI Scientist, Machine Learning Engineer, Research Scientist

Related on AIssential

• Your phone can use tiny skin-colour changes to measure your heart rate — Smartphone cameras can measure heart rate by detecting subtle skin color changes.
• Intervention-Based Self-Supervised Learning: A Causal Probe Paradigm for Remote Photoplethysmography — Intervention-based self-supervised learning can overcome correlation traps in rPPG by actively verifying physical signal hypotheses.
• An Exploratory Study of Blood Glucose Estimation from Photoplethysmography Signals using Machine Learning — Machine learning models can potentially estimate blood glucose non-invasively from smartwatch Photoplethysmogram (PPG) signals.
• Enhancing Construction Worker Safety in Extreme Heat: A Machine Learning Approach Utilizing Wearable Technology for Predictive Health Analytics — Wearable-derived physiological data, analyzed by attention-based LSTMs, can accurately predict heat stress in construction workers.
• How AI is helping improve heart health in rural Australia — AI-driven population health analytics can proactively identify and address health risks in underserved communities.
• CAP: Towards PPG Universal Representation Learning with Patient-level Supervision — CAP improves PPG representation learning by anchoring signal data to patient-level clinical semantics via cross-modal contrastive pretraining.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by AI Magazine.