河北工业科技2025，Vol.42Issue(4)：303-313,11.DOI:10.7535/hbgykj.2025yx04001
基于DRSN-Transformer编码器的域自适应辐射源个体识别方法研究
Research on domain-adaptation specific emitter identification method based on DRSN-Transformer encoder
摘要
Abstract
In order to enable the deep neural networks to accurately identify the emitters of different transmission channels,a domain-adaptation specific emitter identification method based on deep residual shrinkage network(DRSN)fusion Transformer encoder was proposed.DRSN was used to automatically remove the noise from the I/Q received signals through the soft threshold module,and the Transformer encoder was used to further extract the dependent features among symbols in the signals.The domain-adaptation adversarial learning method was used to map target signals from different domains to target features with the same distribution,which enabled the DRSN-Transformer encoder network model to accurately extract the radio frequency fingerprint(RFF)features independent of the channel domain,and to realize the accurate identification of the emitter in the target domain under channel changes.The modulator distortion signal model was used for simulation experiments.The results show that compared with the ResNet model and the DRSN model,the average recognition accuracy of the model in this paper has increased by 2.98 percentage points and 1.65 percentage points respectively.The domain-adaptation adversarial learning method based on DRSN-Transformer encoder network model can effectively reduce the inconsistency of signal feature distribution between the source domain and the target domain,compared with DRSN-Transformer encoder network model trained by traditional methods,the recognition accuracy is increased by 20.73 percentage points at a signal-to-noise ratio of 27 dB,which significantly improves the identification performance when the channel changes.Compared with the traditional learning methods,although the proposed method adds the adversarial training process of the feature extraction network and the domain discrimination network,the final trained feature extraction network can accurately extract fingerprint features that are independent of channel changes,and has certain application value in the specific emitter identification.关键词
计算机神经网络/深度残差收缩网络/Transformer编码器/域对抗神经网络/特定辐射源识别/信道自适应Key words
computer neural network/deep residual shrinkage network(DRSN)/Transformer encoder/domain adversarial neural network(DANN)/specific emitter identification(SEI)/channel adaptation分类
信息技术与安全科学引用本文复制引用
张冠杰,李艳斌,畅鑫,闫红超..基于DRSN-Transformer编码器的域自适应辐射源个体识别方法研究[J].河北工业科技,2025,42(4):303-313,11.基金项目
中国博士后科学基金(2021M693002) （2021M693002）
