基于多特征融合的LPI雷达信号调制样式识别  

Modulation Recognition of LPI Radar Signals Based on Multi-Feature Fusion

作　　者：吴力华 杨露菁[1] 袁园 WU Li-hua;YANG Lu-jing;YUAN Yuan(School of Electronic Engineering,Naval University of Engineering,Wuhan Hubei430033,China;Luoyang Electronic Equipment Test Center,Luoyang Henan 471000,China)

机构地区：[1]海军工程大学电子工程学院,湖北武汉430033 [2]中国洛阳电子装备试验中心,河南洛阳471000

出　　处：《计算机仿真》2023年第5期37-42,93,共7页Computer Simulation

基　　金：国家自然科学基金(41974005)。

摘　　要：针对情报侦察在雷达电子战中的实际应用,在对LPI雷达信号调制样式特点研究的基础上,结合时频分析方法和神经网络分类算法的功能特性,提出了一种基于多特征融合和组合网络(MFCN)的识别方法。首先构建了较为完备的LPI雷达信号仿真模型。其次提取了信号的高斯降噪模糊函数图(GDAFI)和短时傅里叶图(STFTI)等多种特征。最后搭建了融合Resblock、DAE和LIBSVM模块的组合网络,完成LPI雷达信号调制样式的分类识别。仿真结果表明,提出的方法在SNR=-6dB时,对BPSK、Costas、Frank、LFM、T1~T4、P1~P4共12类LPI雷达典型调制样式能达到95%的PSR,并具有较强的稳定性和鲁棒性,相比其它方法具有更好的识别性能。Aiming at the practical application of intelligence reconnaissance in radar Electronic Warfare(EW),a modulation recognition method of Multi-feature Fusion and Combination Network(MFCN)is proposed in this paper,based on the research on the modulation style characteristics of LPI radar signal and functional characteristics of time-frequency analysis method and neural network classification algorithm.First,a relatively complete LPI radar signal simulation model was constructed.And then,the Gaussian Denoising Ambiguity Function Image(GDAFI)and Short Time Fourier Transform Image(STFI)of the signal were extracted.A combined network with Resblock,DAE and LIBSVM modules was built to complete the classification and the modulation recognition of LPI radar signals.The ex-periment results show that,when SNR=-6dB,this method can achieve PSR of 95%for typical modulation styles of 12 types of LPI radar,including BPSK,Costas,Frank,T1 to T4 and P1 to P4,with strong stability and robustness.And compared with other methods,it has better recognition performance.

关 键 词：多特征融合 组合网络 低截获 调制样式 识别 

分 类 号：TN955[电子电信—信号与信息处理]