Unsupervised Specific Emitter Identification: A Multi-Scale Feature Adaptive Fusion Contrastive Learning Algorithm
Jiabao Wang, Guoru Ding, Yutao Jiao, Dongli Zhang, Peng Tang, Guofeng Wei · IEEE Wireless Communications Letters · 2025
Specific Emitter Identification (SEI) plays a vital role in the fields of electromagnetic spectrum management and electronic warfare. However, existing SEI algorithms typically require labeled information, which is often unavailable in non-cooperative communication and untrusted scenarios. To tackle these challenges, we propose an algorithm based on multi-scale feature adaptive fusion contrastive learning (MSFAF-CL) for unsupervised recognition. First, a contrastive learning network framework capable of extracting multi-scale features (MSF) of signals is designed. Then, an adaptive feature fusion (AFF) module is introduced to better assign weights to different features, thereby enhancing classification performance. When constructing positive and negative sample pairs for signals, we analyze the limitations of commonly used signal augmentation techniques and propose an algorithm that uses the signal’s In-phase and Quadrature (I/Q) sequence to construct positive and negative sample pairs directly. Extensive experimental results based on the Automatic Dependent Surveillance-Broadcast (ADS-B) individual signal dataset demonstrate that the proposed algorithm can effectively distinguish individual signals under unsupervised conditions, achieving a high recognition rate on a 15-category signal dataset. Furthermore, the algorithm is more efficient compared with the state-of-the-art algorithms.