[공지사항]을 빙자한 안부와 근황 
Show more

Daily Arxiv

This is a page that curates AI-related papers published worldwide.
All content here is summarized using Google Gemini and operated on a non-profit basis.
Copyright for each paper belongs to the authors and their institutions; please make sure to credit the source when sharing.

Few-Shot Radar Signal Recognition through Self-Supervised Learning and Radio Frequency Domain Adaptation

Created by
  • Haebom

Author

Zi Huang, Simon Denman, Akila Pemasiri, Clinton Fookes, Terrence Martin

Outline

This paper emphasizes the importance of radar signal recognition (RSR) in electronic warfare (EW) and aims to improve the performance of RSR in environments where labeled RF data is scarce. To this end, we propose a self-supervised learning (SSL) method that utilizes masked signal modeling and RF domain adaptation. In a two-step approach, we pre-train a masked autoencoder (MAE) using baseband I/Q signals from various RF domains, and then transfer the learned representations to the radar domain. Empirical results show that a lightweight self-supervised learning ResNet1D model with domain adaptation improves the 1-shot classification accuracy by up to 17.5% (with in-domain signal pre-training) and 16.31% (with out-of-domain signal pre-training) compared to a baseline model without pre-training. In addition, we present reference results for several MAE designs and pre-training strategies, thereby presenting a new benchmark for ultra-low-dataset radar signal classification.

Takeaways, Limitations

•
Takeaways:
◦
We present an effective self-supervised learning method to improve radar signal recognition performance in label-poor environments.
◦
We demonstrate that performance improvements can be achieved by leveraging out-of-domain data via domain adaptation.
◦
The use of lightweight models increases applicability to real-world electronic warfare environments.
◦
We present a new benchmark for classification of ultra-small data radar signals.
•
Limitations:
◦
Further studies are needed to evaluate the generalization performance of the proposed method. More tests are needed on different types of radar signals and electronic warfare environments.
◦
Lack of specific details about the dataset and model used may require review for reproducibility.
◦
Performance evaluation in a real electronic warfare environment is still lacking.
👍
Made with Slashpage