Transformer-Based Modulation Classification Under Impulsive Noise via Adaptive Denoising
IEEE Communications Letters · 2026 · 30 · 2780-2784
DOI: 10.1109/lcomm.2026.3718587 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Communications Letters, 2026. Verfasst von Zirui Yu und 3 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 15 Literaturangaben. Gefördert von: National Natural Science Foundation of China, Beijing Natural Science Foundation.
Abstract (Originalsprache)
Transformer-based architectures have recently been applied to automatic modulation classification (AMC) due to their strong sequence modeling capability. However, under low signal-to-noise ratio (SNR) conditions, transformer-based models often exhibit a clear decline in classification accuracy. This phenomenon can occur when impulsive noise introduces sporadic high-amplitude disturbances, leading to low-SNR scenarios beyond conventional Gaussian noise assumptions. In this letter, a transformer-based adaptive denoising framework is proposed for modulation classification under impulsive noise. Building upon the transformer-based classification module, an adaptive denoising mechanism is incorporated to suppress impulsive interference while preserving the original classification capability of the classification module. Simulation results on the RadioML2016.10A dataset under Bernoulli–Gaussian impulsive noise show that the proposed method effectively improves classification accuracy under different impulsive noise conditions, achieving over 52% accuracy under 5% burst density and outperforming the original Vision Transformer by about 10%, while requiring updates to only 0.94 K parameters (less than 0.015% of the backbone).
Förderung
- National Natural Science Foundation of China
- Beijing Natural Science Foundation
Zitieren
APA
Yu, Z., Zhang, S., Liu, Q., & Zhang, H. (2026). Transformer-Based Modulation Classification Under Impulsive Noise via Adaptive Denoising. IEEE Communications Letters, 30, 2780-2784. https://doi.org/10.1109/lcomm.2026.3718587
Vancouver
Yu Z, Zhang S, Liu Q, Zhang H. Transformer-Based Modulation Classification Under Impulsive Noise via Adaptive Denoising. IEEE Communications Letters. 2026;30:2780-2784. doi:10.1109/lcomm.2026.3718587
BibTeX
@article{Yu2026,
author = {Yu, Zirui and Zhang, Shuhang and Liu, Qingyu and Zhang, Hongliang},
title = {Transformer-Based Modulation Classification Under Impulsive Noise via Adaptive Denoising},
journal = {IEEE Communications Letters},
year = {2026},
volume = {30},
pages = {2780--2784},
doi = {10.1109/lcomm.2026.3718587}
}
EUVIMED