Agent system improves wrist heart rate estimates by selective correction
An Agent Architecture with Tools and Memory for Selective Correction of Wrist PPG Heart Rate Estimates
Machine Learning
Summary
Wrist heart rate sensors often give wrong readings when you move because of motion noise. The authors created a smart system that uses past data and different signals to decide when to fix or keep heart rate readings. This system reduces errors across many datasets and can run in real-time on phones connected to wearables. It also knows when changes might help or hurt the reading, making the results more reliable overall.
What this means in practice
- •For wearable device developers: Integrate an agent that selectively corrects wrist heart rate readings to reduce errors caused by motion artifacts during physical activities.
- •For mobile health app developers: Deploy real-time heart rate correction algorithms on smartphones paired with wearables to improve accuracy without rejecting many data points.
Authors
Xu Wang, Shuhan Yi, Zhen Cheng
Abstract
Wrist photoplethysmography (PPG) heart rate estimates are vulnerable to motion artifacts, and corrections that reduce mean error can worsen individual windows. We propose an agent architecture with tools and memory around a fixed heart rate estimator. Pulse, motion, and limited historical evidence generate candidate estimates, whose predicted benefits and harms guide selection. An independent risk gate determines whether to retain, correct, or reject each output. We trained the architecture on PPG-DaLiA and tested it on Wrist Exercise, GalaxyPPG, and WildPPG with frozen weights. Across these four public datasets, mean absolute error (MAE), macro-averaged over ten estimators, decreased by up to 44.5%, with the largest reduction on GalaxyPPG. Beneficial and harmful corrections decreased and increased absolute error by more than 1 bpm, respectively. Their ten-estimator macro-averaged proportions ranged from 16.06% to 36.64% and from 1.68% to 6.77%, respectively, across the four datasets. At 60% and 80% target coverage, numerical correction on the same accepted windows contributed more to total error reduction than window selection. We implemented real-time wearable acquisition, Bluetooth reception, and smartphone deployment of a selected estimator and the framework. In backend analysis of four prototype recordings from one participant, ten-estimator macro-averaged MAE decreased from 9.09 to 4.45 bpm without additional rejection. Under the evaluated conditions, the architecture improved average accuracy while quantifying correction benefits and harms, supporting wearable heart rate analysis that considers error, correction risk, and coverage.