低截获概率(low probability of intercept,LPI)雷达信号凭借其卓越的抗截获能力,在现代电子战中得到了广泛应用。但LPI雷达信号的低峰值功率使其极易被加性白高斯噪声(additive white Gaussian noise,AWGN)淹没,导致信噪比(signal-to-n...低截获概率(low probability of intercept,LPI)雷达信号凭借其卓越的抗截获能力,在现代电子战中得到了广泛应用。但LPI雷达信号的低峰值功率使其极易被加性白高斯噪声(additive white Gaussian noise,AWGN)淹没,导致信噪比(signal-to-noise ratio,SNR)较低,给信号的检测和识别带来了极大的挑战。为了从AWGN背景中提取原始LPI雷达信号,本文提出了一种名为LPI-U-Net的深度神经网络(deep neural network,DNN),用于端到端的时域LPI雷达信号增强。该网络由特征提取模块(feature extract module,FEM)、特征聚焦模块(feature focus module,FFM)和信号恢复模块(signal recover module,SRM)组成。首先FEM通过卷积操作提取信号的特征,然后FFM利用卷积和通道间注意力进一步关注对信号增强任务有利的特征,最后SRM利用反卷积操作从特征中重构信号,从而完成LPI雷达信号增强。仿真实验表明LPI-U-Net在低SNR下的LPI雷达信号增强性能优于传统信号处理中典型的降噪方法,验证了其可行性和有效性。展开更多
In this paper,we present a novel unimodular sequence design algorithm based on the coordinate descent(CD)algorithm,aimed at countering electronic surveillance(ES)systems based on cyclostationary analysis.Our algorithm...In this paper,we present a novel unimodular sequence design algorithm based on the coordinate descent(CD)algorithm,aimed at countering electronic surveillance(ES)systems based on cyclostationary analysis.Our algorithm not only provides resistance against cyclostationary analysis(CSA)but also maintains low integrated sidelobe(ISL)characteristics.Initially,we derive the expression of the cyclostationary feature(CSF)detector and simplify it into an iterative quadratic form.Additionally,we derive a quadratic form to ensure the similarity of the autocorrelation sidelobes.To balance the minimization of the detection probability and the ISL values,we introduce a Pareto scalar that transforms the multiobjective optimization problem into a convex combination of objective functions.This approach allows us to find an optimal trade-off between the two objectives.Finally,we propose a monotonic algorithm based on the CD algorithm to counter CSA analysis.This algorithm efficiently solves the optimization problem mentioned earlier.Numerical experiments are conducted to validate the correctness and effectiveness of our proposed algorithm.展开更多
以内蒙古巴彦淖尔市磴口县为研究区,在LPI-CA-Markov模型的基础上构建AES-LPI-CA模型,利用人工内分泌系统(AES)调整元胞自动机邻域中中心元胞的转移概率,并基于磴口县2000年与2007年景观格局数据对县域2014年景观格局进行模拟,将该模型...以内蒙古巴彦淖尔市磴口县为研究区,在LPI-CA-Markov模型的基础上构建AES-LPI-CA模型,利用人工内分泌系统(AES)调整元胞自动机邻域中中心元胞的转移概率,并基于磴口县2000年与2007年景观格局数据对县域2014年景观格局进行模拟,将该模型模拟结果与LPI-CA-Markov模型、CA-Markov模型的模拟结果进行对比,结果显示3种模型模拟结果的KIA(Kappa index of agreement,以2014年实际景观分布为参照)依次为0.823 6、0.785 5、0.768 2,AES-LPI-CA模型显示了较高模拟精度。展开更多
针对LPI信号分类识别问题中,时频图像受噪声干扰严重的问题,提出了一种基于二维快速经验模式分解(FBEMD)的图像降噪算法,并利用该算法实现对LPI信号的分类。首先利用时频分析方法,获得待分类信号的时频分布图像;使用二维EMD分解算法对...针对LPI信号分类识别问题中,时频图像受噪声干扰严重的问题,提出了一种基于二维快速经验模式分解(FBEMD)的图像降噪算法,并利用该算法实现对LPI信号的分类。首先利用时频分析方法,获得待分类信号的时频分布图像;使用二维EMD分解算法对图像降噪;截取包含时频信息的图像部分,通过主分量分析法提取特征矢量;最后采用RBF神经网络完成信号的分类识别任务。对常见的LPI雷达信号进行仿真,结果表明较低信噪比情况下,该方法仍能获得较好的分类结果。当信噪比为-2 d B时,采用二维EMD降噪算法,平均正确识别率能够达到93%。展开更多
针对先验信息残缺的非合作电子对抗背景下的低截获概率雷达信号识别问题,提出一种基于改进的半监督朴素贝叶斯的识别算法。该算法首先提取出4种低截获概率(low probability of intercept,LPI)雷达信号的双谱对角切片作为识别特征;针对...针对先验信息残缺的非合作电子对抗背景下的低截获概率雷达信号识别问题,提出一种基于改进的半监督朴素贝叶斯的识别算法。该算法首先提取出4种低截获概率(low probability of intercept,LPI)雷达信号的双谱对角切片作为识别特征;针对传统的半监督朴素贝叶斯(semi-supervised Na?ve Bayes,SNB)在更新训练样本集过程中会产生迭代错误的不足,利用改进的SNB(Revised SNB,RSNB)算法构建分类器,完成对测试样本的识别。该方法通过在无标记样本集生成的置信度列表中选取置信度较高的样本添加到有标记样本集中,再利用预测后的分类结果对分类器参数(即特征期望向量珡mi和方差向量σi)进行改进,有效解决了传统算法分类精度低且分类性能不稳定等缺点。理论分析和仿真结果表明,在LPI雷达信号识别问题,相比于SNB算法和传统的主成分分析加支持向量机法(principal component analysis-support vector machine,PCA-SVM),该算法具有更高的分类识别率和更好的分类性能。展开更多
基金support of the National Natural Science Foundation of China under grant numbers 62101570 and 61901494financial support has played a crucial role in the successful completion of this research.
文摘In this paper,we present a novel unimodular sequence design algorithm based on the coordinate descent(CD)algorithm,aimed at countering electronic surveillance(ES)systems based on cyclostationary analysis.Our algorithm not only provides resistance against cyclostationary analysis(CSA)but also maintains low integrated sidelobe(ISL)characteristics.Initially,we derive the expression of the cyclostationary feature(CSF)detector and simplify it into an iterative quadratic form.Additionally,we derive a quadratic form to ensure the similarity of the autocorrelation sidelobes.To balance the minimization of the detection probability and the ISL values,we introduce a Pareto scalar that transforms the multiobjective optimization problem into a convex combination of objective functions.This approach allows us to find an optimal trade-off between the two objectives.Finally,we propose a monotonic algorithm based on the CD algorithm to counter CSA analysis.This algorithm efficiently solves the optimization problem mentioned earlier.Numerical experiments are conducted to validate the correctness and effectiveness of our proposed algorithm.
文摘以内蒙古巴彦淖尔市磴口县为研究区,在LPI-CA-Markov模型的基础上构建AES-LPI-CA模型,利用人工内分泌系统(AES)调整元胞自动机邻域中中心元胞的转移概率,并基于磴口县2000年与2007年景观格局数据对县域2014年景观格局进行模拟,将该模型模拟结果与LPI-CA-Markov模型、CA-Markov模型的模拟结果进行对比,结果显示3种模型模拟结果的KIA(Kappa index of agreement,以2014年实际景观分布为参照)依次为0.823 6、0.785 5、0.768 2,AES-LPI-CA模型显示了较高模拟精度。
文摘针对LPI信号分类识别问题中,时频图像受噪声干扰严重的问题,提出了一种基于二维快速经验模式分解(FBEMD)的图像降噪算法,并利用该算法实现对LPI信号的分类。首先利用时频分析方法,获得待分类信号的时频分布图像;使用二维EMD分解算法对图像降噪;截取包含时频信息的图像部分,通过主分量分析法提取特征矢量;最后采用RBF神经网络完成信号的分类识别任务。对常见的LPI雷达信号进行仿真,结果表明较低信噪比情况下,该方法仍能获得较好的分类结果。当信噪比为-2 d B时,采用二维EMD降噪算法,平均正确识别率能够达到93%。
