Combining Attention Mechanism with Broad Learning: A Novel Approach to electromagnetic signal Identification

In electromagnetic signal identification, signals from different sources often exhibit only subtle differences in specific features, posing challenges for distinguishing similar signals. To address this issue, this paper proposes a broad learning approach integrated with an attention mechanism, specifically designed for electromagnetic signal identification. The method first constructs and trains a broad learning network to extract features and classify samples, and then embeds a simple attention module between the feature mapping layer and the enhancement layer. This module enables the model to focus on critical features, while its flat structure and pseudoinverse weight computation can reduce computational burdens and improve efficiency.The approach targets key types of electromagnetic signals, including radar signals, communication electromagnetic signals, and electronic warfare electromagnetic signals — all of which require rapid and accurate identification under resource constraints. Experimental results show that the method effectively enhances computational efficiency and model robustness, making it applicable to scenarios such as radar target identification, communication security verification, and electronic warfare countermeasures, where precise differentiation of similar electromagnetic signals is crucial.