闪电瞬态电场信号波形去噪方法  

Denoising methods of lightning transient electrical signals

作　　者：李鹏[1] 郑毅[1] 张义军[2] 

机构地区：[1]防化研究院,北京102205 [2]中国气象科学研究院,北京100081

出　　处：《强激光与粒子束》2007年第12期2055-2059,共5页High Power Laser and Particle Beams

基　　金：国家科技基础条件平台资助课题(2004DEA71070);科技部科技开发项目"三维闪电探测系统研制"资助课题

摘　　要：通过对传统数字滤波和Birge-Massart小波阈值法的比较，对闪电瞬态电场信号进行了去噪研究。构建了双指数衰减脉冲信号并在其上叠加了高斯白噪声信号来仿真闪电瞬态电场实测波形，通过计算去噪前后信号的均方误差和幅值误差，比较了加权平滑滤波、FIR数字低通滤波器和小波阈值法的去噪效果，从处理仿真波形和实测波形结果可以发现，小波阈值去噪方法明显优于另外两种传统滤波方法。通过比较仿真信号由不同小波分解层数去噪结果的均方误差和幅值误差，确定在对闪电电场实测信号的去噪过程中，小波系数分解层数应选5～7。Traditional digital filter method and Birge-Massart threshold wavelet method were used to study the denoisng of lightning transient electrical signals. The mean square error and magnitude error of simulation signals were calculated to compare the denoising effects, and double exponential-decay pulse signals and Gaussian white-nolse were composed as the simulating signals. Denoising simulation and measured signals proved that the Birge-Massart threshold wavelet ways were better than the traditionally used smoothing filter and FIR digital low-pass filter ones. Comparing mean square error and magnitude error with different decomposition orders after simulating signals denoising, it is concluded that wavelet decomposition order in measuring lightning transient electrical signals should be 5 to 7.

关 键 词：闪电瞬态电场信号 波形去噪 小波变换 Birge-Massart阚值法 分解层数 

分 类 号：TM935.4[电气工程—电力电子与电力传动] TN957.1[电子电信—信号与信息处理]