Joint Recognition of Radar Modulations and Composite Jamming via Decoupled Coordinate Attention and Gated Recurrent Units

In cognitive electronic warfare, radar systems must simultaneously perform automatic modulation recognition (AMR) and composite jamming identification (CJI). Most traditional methods employ standard two-dimensional convolutional networks to handle these tasks independently. However, such designs fail to take into account the physical anisotropy of time-frequency representations and the dynamic temporal variations within radar pulses. Moreover, directly combining the two tasks in a simple joint optimization scheme often leads to severe negative transfer. To address this issue, this paper proposes CoGRU-Net, a unified dual-task network that combines Decoupled Coordinate Attention (DCA) and Gated Recurrent Unit (GRU) temporal tracking. The DCA module does not symmetrically fuse features but independently aggregates one-dimensional representations along the time axis and frequency axis, thereby preserving clear directional physical boundaries and eliminating parameter interference across physical domains. Meanwhile, the GRU module models the intra-pulse phase trajectory and captures periodic jamming structures. By leveraging the physical complementarity between radar waveforms and jamming patterns, the two tasks play a mutually regularizing role during training. Comprehensive experiments on a dataset covering 6 modulation types and 17 classes of composite jamming styles demonstrate that the proposed architecture achieved an AMR task accuracy of 97.85%±0.21% and a CJI task accuracy of 98.08%±0.20% across five independent runs (with an overall accuracy rate of 97.97%±0.15%). Under harsh conditions, such as severe thermal noise (SNR∈[−15,−10)dB) and critical spectral aliasing (JSR∈[0,5)dB), the classification accuracy of this model under seed 42 remained above 93.53% and 94.73% respectively. These research results confirmed the effectiveness and robustness of the proposed joint design of spatio-temporal features in a contested electromagnetic spectrum environment.

Authors

Institutions

Publication Details

Journal
Sensors
Published
2026-11-17
DOI
https://doi.org/10.3390/s26227263
Primary Topic
Wireless Signal Modulation Classification
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Joint Recognition of Radar Modulations and Composite Jamming via Decoupled Coordinate Attention and Gated Recurrent Units

Yangyang Wang, Zhiliang Yang, Ran Yan, Yanmei Dou et al.
3 links
Sensors
Wireless Signal Modulation Classification
article

Joint Recognition of Radar Modulations and Composite Jamming via Decoupled Coordinate Attention and Gated Recurrent Units

Yangyang Wang, Zhiliang Yang, Ran Yan, Yanmei Dou, Huiyan Wu
article en

Abstract

In cognitive electronic warfare, radar systems must simultaneously perform automatic modulation recognition (AMR) and composite jamming identification (CJI). Most traditional methods employ standard two-dimensional convolutional networks to handle these tasks independently. However, such designs fail to take into account the physical anisotropy of time-frequency representations and the dynamic temporal variations within radar pulses. Moreover, directly combining the two tasks in a simple joint optimization scheme often leads to severe negative transfer. To address this issue, this paper proposes CoGRU-Net, a unified dual-task network that combines Decoupled Coordinate Attention (DCA) and Gated Recurrent Unit (GRU) temporal tracking. The DCA module does not symmetrically fuse features but independently aggregates one-dimensional representations along the time axis and frequency axis, thereby preserving clear directional physical boundaries and eliminating parameter interference across physical domains. Meanwhile, the GRU module models the intra-pulse phase trajectory and captures periodic jamming structures. By leveraging the physical complementarity between radar waveforms and jamming patterns, the two tasks play a mutually regularizing role during training. Comprehensive experiments on a dataset covering 6 modulation types and 17 classes of composite jamming styles demonstrate that the proposed architecture achieved an AMR task accuracy of 97.85%±0.21% and a CJI task accuracy of 98.08%±0.20% across five independent runs (with an overall accuracy rate of 97.97%±0.15%). Under harsh conditions, such as severe thermal noise (SNR∈[−15,−10)dB) and critical spectral aliasing (JSR∈[0,5)dB), the classification accuracy of this model under seed 42 remained above 93.53% and 94.73% respectively. These research results confirmed the effectiveness and robustness of the proposed joint design of spatio-temporal features in a contested electromagnetic spectrum environment.

SensorsVol. 26(22)
North University of China (CN)
Wireless Signal Modulation Classification
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon