基于GAN的直扩信号生成算法  

DSS signal generation algorithm based on GAN

作　　者：陈丽[1,2] 方梓涵 梅立泉 CHEN Li;FANG Zihan;MEI Liquan(The 54th Research Institute of China Electronics Technology Group Corporation,Shijiazhuang 050081,China;Hebei Key Laboratory of Electromagnetic Spectrum Cognition and Control,Shijiazhuang 050081,China;School of Mathematics and Statistics,Xi’an Jiaotong University,Xi’an 710049,China)

机构地区：[1]中国电子科技集团公司第五十四研究所,河北石家庄050081 [2]河北省电磁频谱认知与管控重点实验室,河北石家庄050081 [3]西安交通大学数学与统计学院,陕西西安710049

出　　处：《系统工程与电子技术》2023年第5期1544-1552,共9页Systems Engineering and Electronics

基　　金：国家自然科学基金(U19B20281009850)资助课题。

摘　　要：将深度学习模型应用至电子干扰技术来生成干扰信号具有重要的现实意义。将生成对抗网络(generative adversarial network,GAN)应用于信号生成领域,对电磁扩频信号频谱数据的分布进行深度学习,并生成与其相干的干扰信号。在实验中GAN的生成器和判别器互相博弈训练,通过自适应矩估计(adaptive moment estimation,Adam)进行优化,最终训练出良好的模型,可以生成所需信号。实验结果表明,基于GAN的信号生成算法生成的数据分布已基本具备真实数据分布普遍具有的特点,对同一信噪比的电磁频谱数据进行深度学习后,生成数据能够较为准确地学习到不同信噪比电磁频谱数据的不同特点。It is of great practical significance to apply deep learning model to electronic jamming technology to generate jamming signals.The generative adversarial network(GAN)is applied to signal generation,and the electromagnetic spread spectrum signal is deeply learned by using the model,and the coherent interference signal is generated by learning the distribution of spectrum data of electromagnetic spread spectrum signal.In the experiment,the generator and discriminator of GAN are trained with each other and optimized by adaptive moment estimation(Adam).Finally,a good model can be trained and the required signals can be generated.Experimental results show that the generated data distribution based on the GAN signal generation algorithm basically has the characteristics of the real data distribution,and the generated data can accurately learn the different characteristics of the electromagnetic spectrum data with different signal to noise ratio(SNR)after deep learning of the same SNR data.

关 键 词：扩频信号 频谱数据 生成对抗网络 电子干扰 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]