Radar Compound Jamming Recognition Based on Image Segmentation and Fused Attention Residual Network

Area of Science:

• Electrical Engineering
• Signal Processing
• Artificial Intelligence

Background:

• Modern electromagnetic environments present complex challenges for radar systems, including sophisticated compound jamming.
• Compound jamming, an additive combination of multiple signal types, is difficult to identify due to its diverse patterns.

Purpose of the Study:

• To develop an effective method for the segmentation and recognition of compound jamming signals.
• To improve the robustness and generalization capability of jamming recognition systems.

Main Methods:

• Utilized image segmentation techniques on time-frequency domain representations (obtained via Short-Time Fourier Transform - STFT) for jamming segmentation.
• Developed an enhanced Residual Network (ResNet) with a spatial-channel fused attention mechanism (SCFAM) for feature extraction and signal recognition.

Main Results:

• Achieved a high recognition accuracy of 98.60% for compound jamming signals.
• Demonstrated superior performance compared to three classical approaches in compound jamming recognition.
• Showcased excellent performance in recognizing untrained compound jamming types, indicating strong robustness and generalization.

Conclusions:

• The proposed method effectively segments and recognizes compound jamming signals in complex electromagnetic environments.
• The integration of attention mechanisms within ResNet enhances the ability to capture multi-level features for improved jamming recognition.
• The method offers a robust and generalizable solution for advanced radar electronic warfare applications.