OpenMHC: Accelerating the Science of Wearable Foundation Models

2026-06-25

AI Summary

This work addresses critical limitations in wearable health research—namely, the scarcity of large-scale public datasets and reproducible foundation models—by introducing OpenMHC, the largest open-source multimodal wearable dataset to date. OpenMHC encompasses over 60 million hours of sensor data from 11,894 participants and uniquely integrates multichannel physiological signals with health behavior variables. Accompanying the dataset, the authors release an open-source wearable foundation model, standardized data processing pipelines, and a unified evaluation benchmark, enabling three core tasks: downstream prediction, data imputation, and time-series forecasting. Through comprehensive benchmarking of both classical and state-of-the-art models—and by fully open-sourcing data, code, and model weights—this study substantially advances open, reproducible, and standardized research in wearable health AI.

Abstract

Mobile and wearable devices offer an unprecedented opportunity for continuous, passive health monitoring and active health coaching. However, the largest wearable datasets are not publicly available for research, and leading wearable foundation models trained on such datasets are rarely open-weight or come with reproducible training code. To accelerate open science in wearable health, we release OpenMyHeartCounts (OpenMHC), the largest and most comprehensive open-access wearable health dataset to date, alongside open-source implementations of recent wearable foundation models. OpenMHC, derived from over a decade of data collected through the My Heart Counts study app, includes>60 million hours of wearable data across 19 sensor channels (e.g., step count, heart rate, sleep, workouts) and up to 169 linked variables, including health, lifestyle, mood, and behavior from 11,894 consenting participants. Furthermore, we introduce a unified, open benchmark that enables standardized comparison of wearable health models across three tracks: health and behavior downstream prediction, multivariate data imputation, and time-series forecasting. We benchmark classical methods alongside recent wearable and multivariate time series foundation models. By open-sourcing data, code, and model weights at this unprecedented scale, we aim to democratize wearable health AI research and enable the community to drive open progress in this domain.

Problem

• wearable foundation models
• open science
• health monitoring
• data accessibility
• reproducibility

Innovation

• wearable foundation models
• open-access dataset
• multivariate time series
• health monitoring
• open benchmark