Node Attention-Embedded Broad Learning Method for Few-Shot Specific Emitter Identification

In non-cooperative communication reconnaissance scenarios, the limited number of intercepted signals introduces the few-shot specific emitter identification (SEI) problem. Deep learning-based SEI methods are prone to overfitting, resulting in poor recognition performance under few-shot conditions. To address this issue, this paper proposes a broad learning-based SEI method embedded with an node attention mechanism. The broad learning system (BLS) is employed to simplify the network structure, mitigating overfitting caused by limited samples, while the node attention mechanism is introduced to enhance the model’s ability to capture important features. Experiments conducted on a publicly available ADS-B dataset demonstrate that BLS effectively addresses the overfitting problem of deep learning under few-shot conditions. Furthermore, due to the incorporation of the node attention mechanism, the proposed method achieves improved recognition accuracy compared to the baseline BLS model.