Age estimation via electrocardiogram from smartwatches

TL;DR

This paper shows that smartwatch ECGs can accurately estimate age, offering a privacy-friendly alternative to traditional methods.

Contribution

A novel smartwatch ECG dataset and improved age estimation accuracy compared to clinical ECG studies.

Findings

Smartwatch ECG age estimation achieved a mean absolute error of 2.93 years.

Accuracy was highest during adolescence due to pronounced ECG changes.

Binary age classification (13–21 years) reached 93–96% accuracy.

Abstract

Age estimation is increasingly vital for regulating access to age-restricted services, especially to protect children online. Traditional methods—ID checks, facial recognition, and databases—raise concerns about privacy and reliability in digital contexts. Electrocardiogram (ECG) signals, reflecting heart activity, offer a promising alternative due to their age-dependent characteristics. However, prior research has largely relied on hospital-grade ECGs, limiting real-world use. To address this, we created a novel data set using smartwatch ECGs from 220 individuals across a broad age range. By testing various features and machine learning models, we achieved a mean absolute error (MAE) of 2.93 years—outperforming clinical ECG-based studies. Accuracy peaked during adolescence, when ECG changes are most pronounced. We also performed binary age classification (13–21 years), reaching 93–96% accuracy. These findings highlight smartwatch ECG’s potential for accurate and privacy-respecting age estimation.

Figures