【摘要】 经验模态分解(EMD)的一个关键问题是处理边界效应。尽管目前除了Huang申请了NASA专利的边界处理方法,仍没有一个最终的解决方案,但工程上已经提出了多种处理方法。本文实现了工程上常用的5种EMD边界处理方法:线性外延,多项式拟合,镜像法,径向基(RBF)神经网络预测和AR预测方法,设计了一套消除了EMD处理中信号的相互作用及模式混淆影响的测试方法,并利用准周期信号和随机信号对它们的边界效应处理结果进行了定量测试。结果表明镜像法是目前相对最优的EMD边界处理方法。更多还原

【Abstract】 One of the most important problems in Empirical Mode Decomposition (EMD) applications is mitigation of the end effect. Except Huang’s patented approach several methods have been proposed. However, a final solution for this problem is yet to be found. In this paper five common end effect mitigation methods of EMD have been investigated, including linear extending method, polynomial fitting extending method, mirror extrema extending method, RBF neural network prediction method and AR prediction method. With a quasi-periodical signal and a stochastic signal as the test bed a quantitative test method was proposed for elimination of the mode confusion effect of EMD. The five end effect mitigation methods were quantitatively evaluated and the comparison shows that mirror extrema extending method is the best option among the five methods.更多还原