Spatio-Temporal Multi-Sensor Anomaly Detection With Concurrent-Noise Distinction
IEEE Signal Processing Letters · 2026 · 33 · 3237-3241
DOI: 10.1109/lsp.2026.3718334 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Signal Processing Letters, 2026. Verfasst von Hua Xiao und 1 weiteren Autor:innen. Die Arbeit nennt 28 Literaturangaben. Gefördert von: National Natural Science Foundation of China, Hubei Provincial International Science and Technology Cooperation.
Abstract (Originalsprache)
Anomaly detection in remote facility monitoring typically relies on multi-sensor signals. However, existing methods often suffer from two major limitations: (1) they fail to distinguish externally induced concurrent noise, namely transient synchronized perturbations shared across multiple sensors but unrelated to the system state, from true anomalies; (2) they prioritize accuracy at the expense of efficiency, making them unsuitable for resource-constrained sensor environments. To tackle these challenges, we propose ${\mathit{STAND}}$ , a two-stage S patio- T emporal A nomaly detection with co N current- N oise D istinction. In the first stage, parallel Temporal Convolutional Networks (TCNs) with dilated causal convolutions are employed to rapidly reconstruct the time series of each sensor. In the second stage, a cross-attention mechanism is introduced to fuse spatio-temporal correlations across monitoring sequences, discerning and suppressing potential concurrent noise among different sensors. This framework facilitates lightweight parallel processing and enables distinction between true anomalies and concurrent noise. Experiments demonstrate that ${\mathit{STAND}}$ achieves competitive performance across multiple public anomaly detection benchmarks in terms of accuracy, recall, F1-score, and AUC. On the AETA dataset, which exhibits significant concurrent noise, the proposed method significantly outperforms existing solutions in processing speed and false alarm rate, effectively ensuring the efficiency and accuracy of remote facility monitoring systems.
Förderung
- National Natural Science Foundation of China
- Hubei Provincial International Science and Technology Cooperation
Zitieren
APA
Xiao, H., & Yuan, J. (2026). Spatio-Temporal Multi-Sensor Anomaly Detection With Concurrent-Noise Distinction. IEEE Signal Processing Letters, 33, 3237-3241. https://doi.org/10.1109/lsp.2026.3718334
Vancouver
Xiao H, Yuan J. Spatio-Temporal Multi-Sensor Anomaly Detection With Concurrent-Noise Distinction. IEEE Signal Processing Letters. 2026;33:3237-3241. doi:10.1109/lsp.2026.3718334
BibTeX
@article{Xiao2026,
author = {Xiao, Hua and Yuan, Jingling},
title = {Spatio-Temporal Multi-Sensor Anomaly Detection With Concurrent-Noise Distinction},
journal = {IEEE Signal Processing Letters},
year = {2026},
volume = {33},
pages = {3237--3241},
doi = {10.1109/lsp.2026.3718334}
}
EUVIMED