Robust stress detection from wearable ECG using subject-specific normalization and stacking ensemble learning

AI Summary

Subject-specific normalization plus multi-domain ECG features (HRV, P-Q-R-S-T morphology, ECG-derived respiration) mitigates individual variability for wearable stress detection.
Stacking ensemble of XGBoost, Random Forest and SVM delivers superior classification performance compared with baseline single models.
Validated on the WESAD dataset, the framework achieves high accuracy and supports robust real-time stress monitoring on personal wearable devices.

Comput Methods Biomech Biomed Engin. 2026 Jul 24:1-10. doi: 10.1080/10255842.2026.2704681. Online ahead of print.

ABSTRACT

Driven by the rising prevalence of mental health issues, this study proposes an automated stress monitoring system using wearable ECG signals. Designed for personal medical applications, the framework overcomes individual physiological variability by combining subject-specific normalization with a multi-domain feature extraction strategy encompassing HRV, P-Q-R-S-T morphology, and ECG-derived respiration (EDR). The classification model utilizes a Stacking Ensemble architecture integrating XGBoost, Random Forest, and SVM. Evaluated on the standard WESAD dataset, the proposed framework achieves high accuracy, significantly outperforming baseline single models. These findings demonstrate its robust capability for real-time stress monitoring on personal wearable devices.

PMID:42497314 | DOI:10.1080/10255842.2026.2704681

Document this CPD