摘要
提出了一种基于拉普拉斯谱图理论的随机游走信号降噪方法。方法首先在图上采用k-NN方法构建邻接图,然后通过核函数理论计算拉普拉斯算子,最后利用隐式欧拉公式求解相应的微分方程,取得了良好的降噪效果。提出的算法避免了现有算法参数多、阈值选取困难、需要很强的前提假设或者大量的先验知识等问题。仿真信号试验表明提出的降噪算法与小波降噪算法、自适应降噪和奇异谱降噪算法相比,能够获得更好的信噪比,具有较好的降噪效果。
A new de-noising method of random walk based on Laplacian spectral graph theory is proposed. Adjacent map is built by k-NN method in the graph. And then,Laplace operator can be calculated by kernel function method. Finally,the implicit Euler＇s formula is used for solving the differential equations. The noise can be effectively reduced. It avoids some disadvantages of traditional de-noising methods,such as many parameters,threshold selection difficult. And,traditional methods usually require strict hypothesis or large amount of prior knowledge. Simulation experiments show that the proposed method can performed well than wavelet noise reduction algorithm,adaptive noise reduction and noise reduction algorithm singular spectrum.
作者
赵晨熙
ZHAO Chen-xi(State Key Laboratory of Shale Oil and Gas Enrichment Mechanisms and Effective Development,Beijing 100101,China;SINOPEC Research Institute of Petroleum Engineering,Beijing 100101,China)
出处
《机械设计与制造》
北大核心
2018年第7期225-228,共4页
Machinery Design & Manufacture
基金
国家科技重大专项"海上油气田关键工程技术"(2016ZX05033-004)
中国石化科技部项目"减震稳扭旋冲钻井提速工具的研制"(P17050-3)联合资助
关键词
拉普拉斯算子
随机游走
降噪
谱图理论
Laplace Operator
Random Walk
De-Noising
Spectral Graph Theory
