摘要
为解决在核爆电磁脉冲(NEMP)探测中经常出现的闪电电磁脉冲(LEMP)干扰问题,提出一种基于广义回归神经网络(GRNN)的识别分类算法。通过对收集到的LEMP与NEMP波形的时域分析,定义9个特征参数用以表征、区分二者波形。通过对GRNN的训练使正地闪、负地闪、云闪、先导四种LEMP波形的识别准确率达到92%以上,LEMP的误报率在万分之二左右,为NEMP探测和闪电检测中的信号分类与识别提供了一种可行的技术途径。
In order to solve the problem of lightning electromagnetic pulse(LEMP)interference in nuclear electromagnetic pulse(NEMP)detection,a recognition and classification algorithm based on generalized regression neural network(GRNN)is proposed.By analyzing the collected LEMP and NEMP waveforms in time domain,nine characteristic parameters are defined to characterize and distinguish the two waveforms.Through the training of GRNN,the recognition accuracy of four LEMP waveforms,such as positive ground flash,negative ground flash,cloud flash and leader,is more than 92%,and the false alarm rate of LEMP is about 2/10,000,which provides a feasible technical way for signal classification and recognition in NEMP detection and lightning detection.
作者
王浩骅
苗家友
朱万华
方广有
WANG Hao-hua;MIAO Jia-you;ZHU Wan-hua;FANG Guang-you(Aerospace Information Research Institute,Chinese Academy of Sciences,Key Laboratoryof Electromagnetic Radiation and Sensing Technology,ChineseAcademy of Sciences,University of Chinese Academy of SciencesSchool of Electronic,Electrical and Communication Engineering,Beijing 100190,China;No.6 Research Institute of rocket Army,Beijing 100094,China;不详)
出处
《核电子学与探测技术》
CAS
北大核心
2021年第1期92-97,共6页
Nuclear Electronics & Detection Technology
基金
国家自然科学基金青年科学基金项目:海洋目标及环境极低频电磁特征与传播机理研究61901441。