类不平衡通信辐射源个体识别

打开文本图片集

中图分类号：TN911.6 文献标志码：A文章编号：1003-3114（2025）04-0844-08

Abstract：Theexistenceofcertaincommunicationemiters，whicharecharacterisedbyashortdurationofuseandhighconcealment，resultsinalimitedamountofdatabeingcolectedforrainingpurposs.Thisleadstotegenerationoftrainingsamplestatare imbalancd，ichiuafftstecuracyofSpecificEmiterIentification（SE）Tddressteisusdentifdodfor identifyingcommunicationemitersbasedonincrementalrandomfeaturetrainingwithaMultivariateLongShort-TemMemoryFuly Convolutional Network（MLST-FCN）isproposedtoenancetheaccracyofSEIinthecontextofclass-imbalance.TheMSTM-FCN modelservesasthefoundationforsignalfeatureextraction，withGausianvectorsgeneratedrandomlyduringthetrainingprocsand splicedtogeterwithtefeaturesextractedbyteetworkforsimultaneous taining.Thisapproachstrengthensodelgeneralizationand improvesitsresiliencetolassbalaneeancingtheoverallacuracyExperimentalsultsemostratethatteproposdetod ismoreeffectiveinaddressingtheissueoflass-imbalaneincommuicationeiters.Furthere，eaccuracyofteproposed methodisconsistentlyigherthanthatofecovetioalmetodevenundervarngdegreesofasmbalanceconditiosisiprovement enhances the accuracy of SEI.

Keywords：MLSTM-FCN;class unbalance;SEI;classification and identificatior

0引言

SEI[1] 又称辐射源“指纹”识别，对接收机接收到的电磁信号进行“指纹”特征提取，根据先验信息对发射此信号的辐射源个体进行分类识别。（剩余9284字）