EMD信号分析方法端点问题的处理

Dealing with end issue of EMD method

经验模态分解(EMD)是一种新的处理非线性、非平稳的数据分析方法,但是在利用样条插值获得上下包络过程中存在着棘手的端点问题。在解决该问题已有的添加极值点算法的基础上,提出了通过添加极值点和对称延拓相结合的方法抑制端点问题的思路和策略。针对一个仿真振动信号,对比分析了直接以数据端点作为极值点、多项式拟合算法、神经网络延拓算法、极值点与对称延拓相结合4种算法的效果,结果显示了所提出方法能有效地抑制端点效应。

EMD（Empirical Mode Decomposition） is a way to process the nonlinear or non-smooth data. But the end issue appears when using spline interpolation to get two envelops of the data. A new method combining ‘symmetrical continuation＇ with ‘adding extremum＇ is proposed for it. A simulated vibration signal is decomposed by four methods for comparison： ‘original end data as one extremum＇, ‘polynomial fitting algorithm＇, ‘neural network continuation algorithm＇ ,and ‘combination of consecutive extremes and symmetrical continuation＇ Results show that the proposed method restrains the end effects effectively.