摘要
针对现有雷达信号分选方法在脉冲丢失、脉冲参差及参数估计误差大等复杂电磁环境下分选性能下降这一不足,提出基于卷积降噪自编码器的雷达信号智能分选方法。该方法将其他脉冲序列视为噪声,目标脉冲序列视为待提取的数据。首先将脉冲序列的到达时间进行编码,并将其转化为二进制编码向量,将编码向量输入卷积降噪自编码器学习目标脉冲序列的内部时间模式,再用训练后的网络对混合脉冲序列进行分选,提取出目标脉冲序列。仿真结果表明,在考虑漏脉冲率、参差脉冲率、TOA估计误差、信噪比等参数变化及存在多功能雷达信号的复杂电磁环境下,该方法的分选正确率均明显优于基于TOA参数的传统方法和使用脉内特征的深度学习方法,证明了该方法的有效性和优越性。
In view of the shortcomings that the existing radar signal sorting methods has low accuracy when in complex electromagnetic environment with more missing pulses or spurious pulses and the large estimation error of parameters,a new intelligent sorting method based on convolutional denoising autoencoder was proposed.This method regarded other pulse sequences as noise,and regarded the target pulse sequence as data to be extracted.Firstly,it encoded the time of arrival pulse sequence,and converted the codes into a two-dimensional image.Then the images were input into the convolutional denoising autoencoder for learning the inner patterns of pulse sequences.Finally,the pulse sequence of interest was extracted from well-trained convolutional denoising autoencoders.The simulation results showed that,considering the parameter changes such as missing pulse rate,staggered pulses,estimation errors of TOA,signal-to-noise ratio and the complex electromagnetic environment with multi-functional radar signals,the sorting accuracy of the proposed method was obviously higher than that of the traditional methods using TOA parameter and the deep learning methods using pulse internal features which proved its effectiveness and superiority.
作者
洪淑婕
孙闽红
王之腾
仇兆炀
HONG Shujie;SUN Minhong;WANG Zhiteng;QIU Zhaoyang(School of Communication Engineering,Hangzhou Dianzi University,Hangzhou 310018,China;Army Engineering University of PLA,Nanjing 210001,China)
出处
《探测与控制学报》
CSCD
北大核心
2022年第5期83-89,96,共8页
Journal of Detection & Control
基金
国家自然科学基金项目(61901149)
国防特色学科发展项目(JCKY2019415D002)。
关键词
信号分选
卷积降噪编码器
脉冲编码
脉冲到达时间
radar signal sorting
convolutional denoising autoencoder
pulse code
time of arrival
