基于改进EVM的雷达PRI调制类型开集识别  

Open-set Recognition of Radar PRI Modulation Type Based on Improved EVM

作　　者：文秋月 王志勇[1] WEN Qiuyue;WANG Zhiyong(School of Mathematical Sciences,University of Electronic Science and Technology of China,Chengdu Sichuan 611731,China)

机构地区：[1]电子科技大学数学科学学院,四川成都611731

出　　处：《现代雷达》2024年第8期22-28,共7页Modern Radar

基　　金：中央高校业务费咨询资助项目(Y030202063010101)。

摘　　要：雷达脉冲重复间隔(PRI)的调制类型是分析雷达工作状态和任务的重要手段。针对常见PRI调制类型识别算法无法识别未知调制类型的问题,文中提出一种基于改进极值机(EVM)的雷达PRI调制类型开集识别方法。首先,采用残差-双向长短时记忆网络进行PRI序列的特征提取;其次,结合原型学习,利用基于距离的交叉熵损失和原型损失对特征提取网络进行训练;最后,在特征空间中引入已知类特征的线性组合以模仿未知类的行为,提出了改进的EVM模型。实验结果表明,与EVM相比,文中所提方法能够提升雷达PRI调制类型的识别准确率,且在开放的电磁环境下具有良好的开集适应性。The modulation type of the radar pulse repetition interval(PRI)is an important means to analyze the working state and task of the radar.In order to solve the problem that common radar PRI modulation type recognition algorithms cannot identify unknown modulation types,an open-set recognition method for radar PRI modulation types based on improved extreme value machine(EVM)is proposed.Firstly,the residual network and bi-directional long short-term memory network is used to extract the features of PRI sequence.Secondly,combined with prototype learning,the feature extraction network is trained by distance-based cross-entropy loss and prototype loss.Finally,an improved EVM model is proposed by introducing a linear combination of known class features into the feature space to simulate the behavior of unknown classes.Experimental results show that compared with EVM,the proposed method can improve the recognition accuracy of radar PRI modulation type,and has good open-set adaptability in open electromagnetic environment.

关 键 词：脉冲重复间隔调制类型 开集识别 残差网络 双向长短时记忆 原型学习 极值理论 

分 类 号：TN957.51[电子电信—信号与信息处理]