改进的一致性正则半监督辐射源个体识别方法  

Improved consistency regularization semi-supervised emitter identification methods

作　　者：吕昊远 俞璐 LYU Hao-yuan;YU Lu(College of Communication Engineering,Army Engineering University of PLA,Nanjing 210007,China)

机构地区：[1]陆军工程大学通信工程学院,江苏南京210007

出　　处：《计算机工程与设计》2022年第3期835-842,共8页Computer Engineering and Design

基　　金：国家自然科学基金项目(61702543)。

摘　　要：针对有标签信号样本数目较少的实际环境中,通信辐射源个体识别技术存在识别准确率较低的问题,提出改进的一致性正则半监督辐射源个体识别方法,在一致性正则方法中引入伪标签思想的改进方案,在3种一致性正则模型上分别加入伪标签正则项。实验中设计适合实采信号数据的Inception深度网络,探究实验参数变化对实验结果的影响,实验结果表明,在通信辐射源个体识别问题上,改进方法比全监督方法、伪标签半监督方法、自编码器特征提取方法和经典的一致性正则半监督方法具有更高的识别准确率。Aiming at the problem of low recognition accuracy in the actual environment with a small number of labeled signal samples,an improved consistency regularization semi-supervised emitter identification method was proposed.The pseudo label idea was introduced into the consistency regularization method,and the pseudo label regular term was added to three consistency re-gularization models.In the experiment,the Inception deep network suitable for the actual signal data was designed,and the inf-luence of experimental parameters on the experimental results was explored.Experimental results show that the improved consistency regularization semi-supervised method has higher recognition accuracy than the full supervised method,pseudo label semi-supervised method,self-encoder feature extraction method and the classical consistency regularization semi-supervised method.

关 键 词：辐射源个体识别 半监督学习 一致性正则 伪标签正则 深度网络 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]