局部特征尺度分解方法及其分量判据研究

Research on Local Characteristic-scale Decomposition and Its Stopping Criteria

在研究内禀时间尺度分解(ITD)方法的基础上提出了一种新的自适应时频分析方法——局部特征尺度分解(LCD)方法,该方法可以自适应地将一个复杂的多分量信号分解为若干个瞬时频率具有物理意义的内禀尺度分量(ISC)之和。对LCD方法的基本原理进行了阐述,并对其分量判据进行了研究,将经验模态分解(EMD)方法中的标准差判据应用于LCD方法。标准差判据的阈值会因自适应时频分析方法的不同而有所差异,因此标准差判据不具有自适应性,针对标准差判据的这一缺陷,提出了一种新的具有自适应性的分量判据——极值单调性判据,该判据无需设定任何阈值。信号分析结果表明了这两种判据的有效性,而极值单调性判据的适用性更强,可直接应用于EMD方法。另外,对比分析了LCD方法和EMD方法的计算效率,分析结果表明LCD方法在计算效率方面要优于EMD方法。

A novel adaptive time--frequency analysis method, LCD was proposed based on the study of ITD. By using the presented method, a complicated multi--component signal could be adap- tively decomposed into a number of intrinsic scale components with physically meaningful instantaneous frequencies. Basic principle of LCD was described in detail after a brief introduction to ITD. To investigate stopping criteria for iterative process,standard deviation criterion was applied to LCD,which was introduced in empirical mode decomposition（EMD）. Varying with the selected adaptive time--frequency analysis method, the threshold of standard deviation criterion had no adaptability. Given the drawback of standard deviation criterion, a new adaptive stopping criterion named extreme monotonicity criterion was proposed,which needed not to set any thresholds. Analysis results of signals demonstrate validity of the two stopping criteria. Moreover,extreme monotonicity criterion gains more applicability than standard deviation criterion and can be also directly applied to EMD. Additionally, computational efficiency of LCD and EMD was investigated,and the comparative analysis results show that LCD is superior to EMD in computational efficiency.