A Knowledge Graph-Driven CNN for Radar Emitter Identification

In recent years, the rapid development of deep learning technology has brought new opportunities for specific emitter identification and greatly improved performance radar identification. The most methods, based on learning, have focused more studying network structures data preprocessing. However, selection utilization a significant impact recognition efficiency, method to adaptively determine two parameters by model yet be studied. This paper proposes knowledge graph-driven convolutional neural (KG-1D-CNN) solve this problem. relationship between is modeled via graph uses 1D-CNN as metric kernel measure these relationships in construction process. process, precise dataset constructed according task requirement. designed recognize target individuals from easy difficult dataset. experiments, algorithms achieved good results high SNR case (10–15 dB), while only proposed could achieve than 90% rate low (0–5 dB). experimental demonstrate efficacy method.