A Time–Frequency-Aware Hierarchical Feature Optimization Method for SAR Jamming Recognition

The recognition of synthetic aperture radar (SAR) jamming has gained significant attention due to its critical role in enhancing radar system performance in complex electromagnetic environments. However, existing research typically addresses a limited number of jamming categories, hindering practical applications that require the simultaneous identification of multiple jamming sources. Moreover, the challenges in feature distribution optimization lead to overlaps in inter-class features and inconsistencies within intra-class features, leading to a decrease in the accuracy of SAR jamming recognition. To tackle these issues, we propose a time-frequency aware hierarchical feature optimization method for SAR jamming recognition. Specifically, we propose a timefrequency aware dual attention mechanism to focus on crucial features in the time-frequency map of SAR jamming signal through parallel attention and unified attention. In addition, we propose a hierarchical jamming feature distribution optimization method that further improves the compactness and discrimination of SAR jamming feature representation. We construct an SAR jamming signal dataset with 50 jamming types to validate the proposed recognition method. The detailed results of ablative experiments and comparative experiments demonstrate the superior performance across different JNR conditions achieved by our method, establishing its efficacy for robust SAR jamming recognition.