为在低复杂度约束条件下提升电磁信号调制识别的性能,提出了一种基于稀疏深度神经网络(Sparse Deep Neural Network,SDNN)的电磁信号调制识别方法。首先,通过提取电磁信号同相和正交两路数据绘制出信号的星座图,作为信号的浅层特征表达...为在低复杂度约束条件下提升电磁信号调制识别的性能,提出了一种基于稀疏深度神经网络(Sparse Deep Neural Network,SDNN)的电磁信号调制识别方法。首先,通过提取电磁信号同相和正交两路数据绘制出信号的星座图,作为信号的浅层特征表达;然后,基于星座图中各信号点密度大小对星座图进行上色,增强星座图中信号特征;最后,通过SDNN对增强后的星座图进行识别分类。实验结果表明,SDNN模型选取合适的剪枝率后,能够有效降低模型存储规模和计算量,其中模型参数压缩了72%,浮点运算量压缩了45%,与原模型97%的综合识别率相比,稀疏化处理后模型的综合识别率为96.8%,在小幅度识别精度损失范围内大幅降低了模型复杂度。展开更多
This paper presents an analytical model for calculating the Earth discontinuous coverage of satellite constellation with repeating ground tracks by integrating and extending the application of coverage region and rout...This paper presents an analytical model for calculating the Earth discontinuous coverage of satellite constellation with repeating ground tracks by integrating and extending the application of coverage region and route theory.Specifically,the visibility condition for a ground point is represented as a coverage region in the two-dimension map of visibility properties,and the trajectories of satellites with circular orbits and repeating ground tracks are converted to several inclined lines in the map.By analyzing the intersections of the lines and the edge of the coverage region,the coverage durations for the ground point can be calculated.Based on the point coverage,the variations of coverage characteristics along the parallel are analyzed,and the regional or global coverage characteristics of constellations can be obtained.Numerical examples show that the proposed method can accurately and rapidly calculate the coverage characteristics,e.g.revisit time and coverage time.The calculated results are extremely close to those of the Satellite Tool Kit(STK)and are also superior to the existing research results.The proposed analytical model can be a useful tool for constellation design and coverage performance analysis.展开更多
随着信息技术的革新和以第五代移动通信技术(5th Generation Mobile Communication Technology,5G)为代表的无线网络技术的快速发展,大量的无线设备需要接入网络。无线网络具有开放式的通信特性,面临着不可忽视的信息安全威胁。射频指...随着信息技术的革新和以第五代移动通信技术(5th Generation Mobile Communication Technology,5G)为代表的无线网络技术的快速发展,大量的无线设备需要接入网络。无线网络具有开放式的通信特性,面临着不可忽视的信息安全威胁。射频指纹技术通过分析无线信号中所携带的能够反映发射机硬件固有缺陷的特征,能够唯一地确定该信号的发射机身份。文章提出利用k均值聚类算法分析差分处理的信号,按照每个聚类的中心点的相位大小顺时针排序,输出分类集群的类别内的距离和类别向量的平均距离,作为分类特征向量和识别目标设备的方法。展开更多
文摘为在低复杂度约束条件下提升电磁信号调制识别的性能,提出了一种基于稀疏深度神经网络(Sparse Deep Neural Network,SDNN)的电磁信号调制识别方法。首先,通过提取电磁信号同相和正交两路数据绘制出信号的星座图,作为信号的浅层特征表达;然后,基于星座图中各信号点密度大小对星座图进行上色,增强星座图中信号特征;最后,通过SDNN对增强后的星座图进行识别分类。实验结果表明,SDNN模型选取合适的剪枝率后,能够有效降低模型存储规模和计算量,其中模型参数压缩了72%,浮点运算量压缩了45%,与原模型97%的综合识别率相比,稀疏化处理后模型的综合识别率为96.8%,在小幅度识别精度损失范围内大幅降低了模型复杂度。
基金the National Natural Science Foundation of China (No. 12072365)the Hunan Provincial Natural Science Foundation of China (No. 2020JJ4657)
文摘This paper presents an analytical model for calculating the Earth discontinuous coverage of satellite constellation with repeating ground tracks by integrating and extending the application of coverage region and route theory.Specifically,the visibility condition for a ground point is represented as a coverage region in the two-dimension map of visibility properties,and the trajectories of satellites with circular orbits and repeating ground tracks are converted to several inclined lines in the map.By analyzing the intersections of the lines and the edge of the coverage region,the coverage durations for the ground point can be calculated.Based on the point coverage,the variations of coverage characteristics along the parallel are analyzed,and the regional or global coverage characteristics of constellations can be obtained.Numerical examples show that the proposed method can accurately and rapidly calculate the coverage characteristics,e.g.revisit time and coverage time.The calculated results are extremely close to those of the Satellite Tool Kit(STK)and are also superior to the existing research results.The proposed analytical model can be a useful tool for constellation design and coverage performance analysis.
文摘随着信息技术的革新和以第五代移动通信技术(5th Generation Mobile Communication Technology,5G)为代表的无线网络技术的快速发展,大量的无线设备需要接入网络。无线网络具有开放式的通信特性,面临着不可忽视的信息安全威胁。射频指纹技术通过分析无线信号中所携带的能够反映发射机硬件固有缺陷的特征,能够唯一地确定该信号的发射机身份。文章提出利用k均值聚类算法分析差分处理的信号,按照每个聚类的中心点的相位大小顺时针排序,输出分类集群的类别内的距离和类别向量的平均距离,作为分类特征向量和识别目标设备的方法。
