E-SDHGN: A Multifunction Radar Working Mode Recognition Framework in Complex Electromagnetic Environments

Abstract

A multifunction radar (MFR) can operate in multiple modes and perform various tasks such as surveillance, detection, fire control, search and tracking. Recognising an MFR's operating mode is critical in electronic warfare and intelligence reconnaissance, aiding practical threat assessment and countermeasure tasks. However, current recognition methods face challenges such as overlapping parameters among working modes and suboptimal recognition accuracy under conditions with parameter errors, missing pulses and false pulses. Spurred by these concerns, this paper proposes an entropy-enhanced spatial-deformable hybrid multiscale group network (E-SDHGN) to recognise the operating mode of an MFR and address these challenges. E-SDHGN employs multidimensional entropy computations to construct robust features and integrates deformable convolution and positional encoding to enhance the model's ability to capture complex features. Additionally, it enhances feature extraction and fusion within the dynamic shared residual network (DSRN) module by integrating KAN modules and hybrid weight-sharing strategies. Additionally, an adaptive margin feature module based on attention mechanisms improves classification accuracy in overlapping parameter conditions. Experimental results demonstrate that E-SDHGN achieves superior recognition accuracy and robustness, even under challenging parameter errors, missing pulses and false pulses. This underscores its value for applications in complex electromagnetic environments.