@article{ART003332387},
author={JaeHyuk, Lee},
title={An Empirical Study on a Domain Adaptation Framework for Personalized Exercise Intensity Prediction Using Smartwatch Sensing},
journal={Journal of Internet of Things and Convergence},
issn={2466-0078},
year={2026},
volume={12},
number={2},
pages={79-85}
TY - JOUR
AU - JaeHyuk, Lee
TI - An Empirical Study on a Domain Adaptation Framework for Personalized Exercise Intensity Prediction Using Smartwatch Sensing
JO - Journal of Internet of Things and Convergence
PY - 2026
VL - 12
IS - 2
PB - The Korea Internet of Things Society
SP - 79
EP - 85
SN - 2466-0078
AB - This study investigated cross-subject generalization in smartwatch-based exercise intensity classification and proposed a lightweight domain adaptation framework under constrained sampling conditions. Exercise intensity was defined relative to individual peak heart rate (%HR_peak) using wrist photoplethysmography and inertial signals. A lightweight 1D-CNN was trained and evaluated using a subject-wise hold-out design to assess cross-subject domain transfer. An unsupervised test-time batch normalization (BN) recalibration strategy was applied to mitigate distribution mismatch. The model achieved strong validation performance (macro-F1 = 0.88, accuracy = 0.87). However, macro-F1 declined to 0.31 under cross-subject evaluation, indicating substantial degradation in detecting moderate- and high-intensity segments. BN adaptation improved macro-F1 to 0.35 and enhanced moderate-intensity detection. These findings demonstrate that cross-subject domain shift fundamentally constrains model generalization and that normalization-based lightweight adaptation can partially mitigate distribution mismatch. The proposed framework provides a scalable foundation for robust model development in wearable-based exercise monitoring.
KW - Cross-subject generalization;Domain adaptation;Exercise intensity prediction;Wearable sensing;1D-CNN
DO -
UR -
ER -
JaeHyuk, Lee. (2026). An Empirical Study on a Domain Adaptation Framework for Personalized Exercise Intensity Prediction Using Smartwatch Sensing. Journal of Internet of Things and Convergence, 12(2), 79-85.
JaeHyuk, Lee. 2026, "An Empirical Study on a Domain Adaptation Framework for Personalized Exercise Intensity Prediction Using Smartwatch Sensing", Journal of Internet of Things and Convergence, vol.12, no.2 pp.79-85.
JaeHyuk, Lee "An Empirical Study on a Domain Adaptation Framework for Personalized Exercise Intensity Prediction Using Smartwatch Sensing" Journal of Internet of Things and Convergence 12.2 pp.79-85 (2026) : 79.
JaeHyuk, Lee. An Empirical Study on a Domain Adaptation Framework for Personalized Exercise Intensity Prediction Using Smartwatch Sensing. 2026; 12(2), 79-85.
JaeHyuk, Lee. "An Empirical Study on a Domain Adaptation Framework for Personalized Exercise Intensity Prediction Using Smartwatch Sensing" Journal of Internet of Things and Convergence 12, no.2 (2026) : 79-85.
JaeHyuk, Lee. An Empirical Study on a Domain Adaptation Framework for Personalized Exercise Intensity Prediction Using Smartwatch Sensing. Journal of Internet of Things and Convergence, 12(2), 79-85.
JaeHyuk, Lee. An Empirical Study on a Domain Adaptation Framework for Personalized Exercise Intensity Prediction Using Smartwatch Sensing. Journal of Internet of Things and Convergence. 2026; 12(2) 79-85.
JaeHyuk, Lee. An Empirical Study on a Domain Adaptation Framework for Personalized Exercise Intensity Prediction Using Smartwatch Sensing. 2026; 12(2), 79-85.
JaeHyuk, Lee. "An Empirical Study on a Domain Adaptation Framework for Personalized Exercise Intensity Prediction Using Smartwatch Sensing" Journal of Internet of Things and Convergence 12, no.2 (2026) : 79-85.
