中文摘要：

经验模态分解（Empirical mode decomposition,EMD）自提出以来已广泛用于信号检测与处理,但其存在很多的缺陷,如频率混叠现象等。为了减轻混叠现象,提取真实的频率成分,本文分析了信号及其一阶导数和二阶导数的关系,作出一种用信号二阶导数的极值点处的信号值取代原EMD算法中的信号极值点进行三次样条插值的方法,其余计算流程不变,仍采用镜像拓延法改善端点效应。仿真结果与原EMD计算结果对比表明,基于信号二阶导数改进的算法能准确分解出信号中幅值分量表现不明显的高频信号,具有实用价值。

英文摘要：

Empirical mode decomposition （EMD） has lots of defects, such as the frequency aliasing phenomena. To reduce the aliasing, it is helpful to analyze the relationship between the signal and its first-order and second-order derivatives. Consequently, a modified empirical mode decomposition （MEMD） interpolation points algorithm is proposed, based on the extremes of the second-order derivative of the signal. Experiment represented that the derivative of signal can clearly characterize the frequency characteristics of high frequency component whose amplitude characteristic is not obvious in the mixed signal. The comparative experiments show that MEMD extract the high frequency, and small value components form signal is better and more accurate than the original EMD.