基于非参数模型的雷达辐射源智能识别  

A Non-Parametric Model-Based Intelligent Recognition Method for Radar Emitters

作　　者：任浩浩 马寒菲 魏裕璐 余思慧 周云 REN Haohao;MA Hanfei;WEI Yulu;YU Sihui;ZHOU Yun(School of Information and Communication Engineering,University of Electronic Science and Technology of China,Chengdu 611731,Sichuan,China;Shanghai Academy of Spaceflight Technology,Shanghai 201109,China)

机构地区：[1]电子科技大学信息与通信工程学院,四川成都611731 [2]上海航天技术研究院,上海201109

出　　处：《空天防御》2024年第5期36-44,共9页Air & Space Defense

基　　金：国家自然科学基金项目(62201124)。

摘　　要：针对现代错综复杂电磁环境中辐射源数量增多、脉冲流密度增大等问题,为实现雷达辐射源高精度智能识别,本文提出一种基于非参数模型的雷达辐射源智能识别方法,该方法主要包含辐射源静态聚类、辐射源类代表点选取和面向数据流的增量分选3个关键步骤。与传统聚类分析方法相比,本文采用Dirichlet过程混合模型的聚类思想,可自适应地完成雷达辐射源类别划分。为在降低数据存储压力的同时最大化保留类分布属性表达,首先,建立基于梯度准则的辐射源类代表点选择策略;其次,针对流式数据增量处理的需求,提出基于标签传播的类别推理方法;最后,通过仿真模拟两种场景的雷达辐射源数据,对比7种聚类算法,验证了本文算法具有较好的识别准确率和鲁棒性。Given the problem of increasing the number of radar emitters and pulse density in a current complex electromagnetic environment,this study proposed an intelligent recognition algorithm using a non-parametric model to achieve high-precision radar emitters recognition.The proposed method was composed of three vital steps:static clustering of radar emitters,selection of representative points for radar emitter classes,and oriented-data stream incremental recognition.Compared to conventional clustering analysis methods,the proposed method utilized the clustering idea of Dirichlet Process Mixture Models(DPMM),which adaptively realized radar emitters’class sorting.To reduce the data storage burden while maximizing the preservation of class distribution,a gradient-based representative point selection strategy for emitter classes was established.Besides,to satisfy the requirement of incremental processing of the data stream,a label propagation-based classification inference method was applied to achieve incremental radar emitters recognition.Finally,the proposed algorithm was validated using simulated radar emitters’data in two scenarios.The evaluation results show that the proposed method is more robust and accurate than the other seven advanced emitters recognition clustering algorithms.

关 键 词：雷达辐射源识别 数据流聚类 非参数模型 

分 类 号：TN957[电子电信—信号与信息处理]