WiTWS: WiFi CSI Path Separation for Calibration-Free Through-Wall Human Activity Sensing
IEEE Transactions on Mobile Computing · 2026-09 · 25 (9) · 14580-14591
DOI: 10.1109/tmc.2026.3682297 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Transactions on Mobile Computing, 2026-09. Verfasst von Xingcan Chen und 3 weiteren Autor:innen. Die Arbeit nennt 34 Literaturangaben. Gefördert von: National Natural Science Foundation of China, Central Guided Local Science and Technology Development Special, USTB POC Project.
Abstract (Originalsprache)
WiFi channel state information (CSI)-based through-wall human activity recognition (HAR) approaches face some fundamental limitations, such as multi-path entanglement obscures human-related propagation components, position-sensitive features cause activity misclassification under spatial variability, and wall heterogeneity induces domain shifts that degrade model generalizability. To overcome these challenges, we propose a novel WiFi CSI path separation-based through-wall sensing approach (WiTWS). Specifically, WiTWS first isolates human-related components by differential phase decomposition with spatial steering vectors to eliminate static and dynamic interferences. WiTWS then introduces an electromagnetic-informed time-frequency adaptive filter leveraging wavelet transforms to cancel wall-specific distortions without prior material knowledge. After that, WiTWS derives direction-change ratios as position-independent features that encode motion geometry, which decouples activity-related features from transceiver placement. These purified features are finally fed into a lightweight spatio-temporal feature fusion network (LSTFFN) for efficient classification. Extensive experiments in real-world scenarios demonstrate that WiTWS achieves an average accuracy of 99.23% in training domains and maintains 97.13% in unseen environments. The proposed approach achieves state-of-the-art through-wall HAR performance without site-specific calibration, and can be robustly extended to unknown through-wall scenarios.
Förderung
- National Natural Science Foundation of China
- Central Guided Local Science and Technology Development Special
- USTB POC Project
Zitieren
APA
Chen, X., Li, C., Meng, W., & Xiao, W. (2026). WiTWS: WiFi CSI Path Separation for Calibration-Free Through-Wall Human Activity Sensing. IEEE Transactions on Mobile Computing, 25(9), 14580-14591. https://doi.org/10.1109/tmc.2026.3682297
Vancouver
Chen X, Li C, Meng W, Xiao W. WiTWS: WiFi CSI Path Separation for Calibration-Free Through-Wall Human Activity Sensing. IEEE Transactions on Mobile Computing. 2026;25(9):14580-14591. doi:10.1109/tmc.2026.3682297
BibTeX
@article{Chen2026,
author = {Chen, Xingcan and Li, Chenglin and Meng, Wei and Xiao, Wendong},
title = {WiTWS: WiFi CSI Path Separation for Calibration-Free Through-Wall Human Activity Sensing},
journal = {IEEE Transactions on Mobile Computing},
year = {2026},
volume = {25},
number = {9},
pages = {14580--14591},
doi = {10.1109/tmc.2026.3682297}
}
EUVIMED