Samsung Advances Wearable Health AI

Samsung Research is developing artificial intelligence models designed to interpret health data from wearable devices and generate continuous, personalized insights. The company says the work could support more preventive and connected approaches to digital health by analyzing signals such as heart rate, sleep and physical activity.

Researchers at Samsung Research America recently introduced two foundation models based on wearable data: xMAE and HiMAE. The studies were accepted by the International Conference on Machine Learning and the International Conference on Learning Representations, respectively.

The xMAE model is designed to learn relationships between electrocardiogram (ECG) and photoplethysmography (PPG) signals. By using continuously collected PPG data to reconstruct portions of ECG signals, the model aims to identify cardiovascular features without requiring users to take separate, active ECG measurements. Samsung said xMAE was trained on about 9,400 hours of ECG and PPG data and outperformed other models in 15 of 19 evaluation tasks.

HiMAE analyzes wearable signals across multiple time scales, allowing it to identify patterns linked to functions such as heart rate, sleep and physical activity. Samsung said the model can operate on a smartwatch-class processor in less than one millisecond, raising the possibility of real-time health analysis on devices without relying on cloud servers.

The company presented the research as part of its Connected Care vision, discussed at the Health Forum during Galaxy Unpacked in July 2026. Samsung said the models are intended to support a range of future health applications, including biosignal analysis, biomarker development and health-risk prediction.