Active Domain Adaptation for mmWave-based HAR via Renyi Entropy-based Uncertainty Estimation
Mingzhi Lin, Teng Huang, Han Ding, Cui Zhao, Fei Wang, Ge Wang, Wei Xi
TL;DR
mmADA tackles the critical issue of domain shift in mmWave-based HAR by introducing a Renyi entropy-based active domain adaptation framework. It combines Evidential Deep Learning with a Renyi-entropy uncertainty selector, pseudo-labeling via a Pseudo Label Set, and contrastive learning to efficiently adapt from a labeled source to unlabeled target data using a small labeling budget. The approach yields state-of-the-art cross-domain accuracy (over 90% in cross-user, cross-position, and cross-environment evaluations) and demonstrates strong generalization on unseen users, environments, and large open datasets like XRF55 and MM-Fi. This work provides a practical, data-efficient pathway for deploying mmWave HAR systems in diverse real-world settings, where labeling costs are prohibitive and domain shifts are pervasive.
Abstract
Human Activity Recognition (HAR) using mmWave radar provides a non-invasive alternative to traditional sensor-based methods but suffers from domain shift, where model performance declines in new users, positions, or environments. To address this, we propose mmADA, an Active Domain Adaptation (ADA) framework that efficiently adapts mmWave-based HAR models with minimal labeled data. mmADA enhances adaptation by introducing Renyi Entropy-based uncertainty estimation to identify and label the most informative target samples. Additionally, it leverages contrastive learning and pseudo-labeling to refine feature alignment using unlabeled data. Evaluations with a TI IWR1443BOOST radar across multiple users, positions, and environments show that mmADA achieves over 90% accuracy in various cross-domain settings. Comparisons with five baselines confirm its superior adaptation performance, while further tests on unseen users, environments, and two additional open-source datasets validate its robustness and generalization.