【作者】 刘旭；

【导师】 罗鹏飞； 李纲；

【作者基本信息】 国防科学技术大学 ， 信息与通信工程， 2011， 硕士

【摘要】 雷达辐射源识别是指对侦收到的雷达信号进行特征提取，并根据先验信息确定发射信号的雷达个体的过程。由于雷达辐射源识别在军事侦察中所占据的独特地位，使得其成为了当今雷达对抗领域的研究热点。目前，雷达辐射源识别的研究已不少，但取得突破性的进展不多。大多从现象去进行无意调制特征的提取，缺少对无意调制的机理以及无意调制在信号中表现形式的研究，使得提取无意调制特征有一定的盲目性；由于无意调制特征极其细微，使得其易受环境的影响，如何方便高效地提取稳健的无意调制特征一直是雷达辐射源识别的难点。论文从雷达发射机频率源的无意调制机理分析入手，分析了主振放大式发射机频率源每一部件产生相位噪声的机理。同时通过研究各部件之间相位噪声的传递特性，得出了每个部件的相位噪声分布以及含有源滤波器的雷达发射机频率源最终输出相位噪声的分布。然后依据之前的分析结果，借助仿真工具ADs2009对主振放大式发射机频率源进行建模，研究发射机频率源每一部件产生的相位噪声以及对总的输出相位噪声的影响，并得到频率源输出的总相位噪声的分布情况。再把仿真得到的输出总相位噪声加入到理想常规雷达信号中以研究细微相位噪声对信号的影响。仿真结果表明，小的附加相位噪声对信号的相位、功率谱以及双谱具有显著改变，证实了相位噪声作为无意调制确实存在于雷达信号之中且本身具有一定的分布特性，同时也说明了提取无意调制的可能性。最后基于脉冲包络的形状能够反映雷达辐射源的个体特征，定义了三种新的包络特征，即拟合上升角、拟合下降角和拟合边沿周长面积比，通过仿真证实，新定义的包络特征较传统的包络特征具有更好的聚类效果，新的包络特征与旋转信号的统计特征及信号功率谱的相关特征一起构成一个新的特征向量，通过支持向量机分类器对信号进行分类：仿真实验表明，在不同信噪比情况下，新的特征向量具有更优的分类效果；实测数据也表明利用新特征向量的识别算法针对同型号、同工作模式的雷达辐射源亦能具备较好的分类效果。更多还原

【Abstract】 Radar Emitter Identification(REI)is to determine radar individual，usingextractingfeatures of received radar signals，by comparingthosefeatureswith a priorinformationlibrary．AsREI plays allimportant roleinthemilitary surveillance，ithasbecomethefoCUSofradarcountenneaanres research．Though the research on REI has been increasing recently，there are fewbreakthrough progresses．Most of the methods extract unintentional modulation onpulse(UMOP)feature just from signal phenomenon．Lacking of research on themechanism of UMOP or the form of unintentional modulation in signal，these makeextraction ofunintentional feature blind．Because UMOP is very subtle，its feature isvulnerabletoenvironmentalimpactsanditisdifficultto extractthe robustfeatures ofUMOP allthetime．We investigate the mechanism of generating PN in each component of MasterOscillator Power-Amplifier(M．0．P．A)transmitter frequency source from the analysisonUMOP．PN ofeach componentandthetotalPNoftheloopincludedactivefilterarerealized by studying the PN transferring characteristics of each component．Then，basedon the previous analysis，simulation tool ADS2009 is utilized to model the M．O．P．Atransmitter frequency source and investigate the PN charac~fistic and the effects ontotal PN of each component in detail，and then the total output PN distribution oftheloop ale obtained．Inorderto studythe effects ofsubtlePN，theoutputofthetotalPNisadded to ideal conventional radar signal．The simulation results indicate that there aredistinctdifferencesbetweenthe original signalandthe signal addedwithphasenoiseinthe phase，power spectrum and bispectrum although the added phase noise is very subtle．It can provide possibility to extract the feature ofUMOP．Finally，as the shape ofthepulse envelope can reflect the individual characteristics of radar emitter，three newfeatures，thefittingupangle，fittingdownangle aswell asthe ratio ofperimeterto area，are defined．Simulation results show that the new envelope features has betteraggregation than the traditional characteristics．Also the mean，the variance，theskewness，the kurtosisandtheenergyentropy ofthe proposed rotation components byintrinsictime-scaledecompositionareutilizedasthenovelfeaturefactorforthe suppoavector machine(SVM)classifier．The characteristic sets including envelope features ofthe new definition has better classification results atdifferent si~alto noise ratio．Allthe results are verified with the real measurement data generated by the uniform typeradar under the same way of work．It shows that a higher recognition rate could beobtained更多还原