Specific Emitter Identification (SEI) of ADS-B Signals via Multi-Head Attention Mechanism

To address the issue of low recognition accuracy in ADS-B signal-specific emitter identification (SEI), a novel method based on the multi-head attention mechanism (MHAM) is proposed. This method takes ADS-B signals,extracting steady-state features at low frequencies using multi-scale wavelet decomposition, and extract transient features from time-frequency information for stitching and fusion. The fused feature dataset is trained using a multi-head attention mechanism to make the results more convincing and practical. The experimental results show that after using (convolutional neural network) CNN-multi-head attention fusion network, the proposed ensemble method achieves a 6%–22% relative improvement in accuracy over a standalone neural network baseline. At the same time,the algorithm has a identification rate of 93% under 30db conditions and an 82% identification rate under 10db conditions, which verifies its good robustness. The algorithm in this article effectively meets the adaptability of multi features and network identification.