Samsung trains AI to detect health patterns hidden in wearable data

Samsung Research on Friday detailed two artificial intelligence models designed to identify health patterns in data collected by wearable devices, as the technology company expands its research into continuous health monitoring.

The models, called xMAE and HiMAE, are designed to analyze so-called biosignals — measurements such as heart activity, movement and sleep-related data collected by devices including smartwatches.

The aim is to help AI recognize patterns across these signals rather than treating each measurement separately, Samsung said.

The announcement follows Samsung’s broader push into AI-powered health technology. In July, the company launched a beta version of Samsung Health Assistant, which uses users’ health data to provide personalized wellness recommendations.

Samsung also presented its broader “Connected Care” vision, which seeks to use AI and wearable data for more continuous and personalized health monitoring.

The two models are research projects rather than newly launched medical products. xMAE focuses on relationships between different heart-related signals, while HiMAE is designed to identify patterns across different time scales, from rapid changes such as heartbeats to longer-term patterns linked to sleep and activity.

Samsung said HiMAE can operate on a smartwatch-class processor in less than a millisecond.

The work is part of a broader effort to develop health foundation models — AI systems trained to recognize patterns in large amounts of biological data and then adapted for different tasks.