94% of researchers rate our articles as excellent or good
Learn more about the work of our research integrity team to safeguard the quality of each article we publish.
Find out more
ORIGINAL RESEARCH article
Front. Signal Process.
Sec. Signal Processing Theory
Volume 5 - 2025 | doi: 10.3389/frsip.2025.1567926
Deep Learning-Based GNSS Composite Jamming Detection and Recognition Technology
Provisionally accepted- School of Electronic Science, National University of Defense Technology, Changsha, China
As the interference environment of Global Navigation Satellite Systems (GNSS) becomes increasingly complex and diverse, real-time and precise interference detection and identification technologies are crucial for enhancing the anti-interference capabilities of receivers. However, most existing interference detection and identification methods focus on single interference types, with limited research on composite interference and a lack of quantitative conclusions. Therefore, this study investigates composite interference detection and identification techniques using deep learning methods, improving the system's capability to detect and identify composite interference. This paper first constructs single interference model and composite interference model, proposes three signal preprocessing methods, and generates corresponding image datasets. Subsequently, the interference detection and identification performance under different signal preprocessing methods is analyzed using the ResNet-18 deep learning neural network. The optimal signal preprocessing method is identified, and quantitative conclusions are obtained. Finally, a lightweight network, LcxNet-Fusion, is designed, which significantly reduces the number of network parameters and forward processing time while maintaining an acceptable level of accuracy reduction. Results show that among the timefrequency 2D diagrams, power spectral diagrams, and histograms generated by signal preprocessing, the time-frequency diagram yields the best detection and identification performance. When the detection rate reaches 90%, the jamming-to-noise ratio (JNR) sensitivity of the time-frequency diagram is -20 dB; when the identification rate reaches 90%, the JNR sensitivity of the time-frequency diagram is -13 dB. On the Tesla V100 GPU, the LcxNet-Fusion network has 24.32 MB parameters, an 43% reduction compared to the ResNet-18 network, with a forward processing time of 1.25 seconds,
Keywords: Composite interference, Detection and recognition, deep learning, Signal preprocessing, Lightweight
Received: 20 Feb 2025; Accepted: 19 Mar 2025.
Copyright: © 2025 Liu, Ren, Xie and Chen. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
* Correspondence:
Feiqiang Chen, School of Electronic Science, National University of Defense Technology, Changsha, China
Disclaimer: All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.