| 摘要: |
| 为了快速高效地实现信号的本征模态分解, 消除分解过程中边界效应, 在Directly-Mean EMD 方法的基础上, 通过引入左、右待估中点和左、右延拓中点, 并根据信号首尾两端各种可能的情况给出了四中点估计公式, 建立了可完全消除边界效应的Directly-Mean EMD 四中点估计方法。该方法不仅减少了样条插值次数, 提高分解速度, 而且还可以有效避免因插值节点过于稀疏所产生的大幅波动, 使分解结果更加准确。将新方法应用于日长数据序列的本征模态函数分解, 得到了满意的分解结果。 |
| 关键词: 本征模态函数 Directly-Mean EMD 方法 四中点估计方法 希尔伯特－黄变换 日长(LOD) |
| DOI： |
| 分类号: |
| 基金项目:国家自然科学基金资助项目(40876094) |
|
| Four-midpoint estimation method removing boundary effect of EMD |
|
|
| Abstract: |
| On the basis of directly-mean EMD method, four-midpoint estimation formula according to all kinds of possible cases at both ends of signal was given by introducing the concepts of left and right undetermined middle points, left and right extended middle points. Four-midpoint estimation method for EMD, which can eliminate boundary effect completely, was established. This new method decompose signal into intrinsic mode functions and eliminate boundary effect more quickly and efficiently. The improvement on decomposition speed was brought by the reduction of the spline interpolation number. And the large fluctuation generated due to too sparse interpolation nodes was avoid, which made decomposition results more accurate. The new method was applied to the mode decomposition of length of day (LOD) data and a satisfactory decomposition result is obtained. |
| Key words: intrinsic mode function (IMF) directly-mean EMD method four-midPoint estimation method Hilbert-Huang Transform LOD (length of day) |
