遗传算法在EMD虚假分量识别中的应用  

Applying Genetic Algorithms in EMD False Component Identification

作　　者：宋娜[1] 石玉[1] 周克印[1] 

机构地区：[1]南京航空航天大学自动化学院

出　　处：《计量学报》2015年第4期413-417,共5页Acta Metrologica Sinica

基　　金：国家科技支撑计划(2012BAA01B00);中央高校基本科研业务费专项资金(NS2012090)

摘　　要：针对EMD（经验模态分解）产生虚假分量这一问题，将遗传算法和K—L散度相结合，对虚假分量进行研究。该方法是先将原始信号进行EMD得到固有模态分量（IMF）；将遗传算法和基于均方积分误差的窗宽最优化原则相结合，分别对原始信号和各个IMF分量优化选取窗宽；然后运用核密度估计方法分别得到它们的概率密度函数估计；最后计算原始信号与IMF分量之间的K-L散度值，设定K—L阈值，将K-L散度值大于阈值的IMF分量去除。实验证明，该方法能准确而又快速地获得实验数据的窗宽，虚假成分与真实分量的K—L值有明显差别，根据设定的阈值能准确识别虚假分量。As EMD （Empirical Mode Decomposition） produces false component, the false component was studied by combining genetic algorithm with Kullback-Leibler divergence. First, the original signal was decomposed into several intrinsic mode functions （IMF）; the original signal and each IMF component were respectively selected the optimal bandwidth that the genetic algorithm and the optimization principles of bandwidth based on integral mean square error were combined;and then applied kernel density estimation methods to get their probability density function estimation;Finally, the Kullback-Leibler divergence between the original signal and each IMF was calculated, setting the threshold of K-L divergence,IMF component whose K-L divergence is greater than the threshold can be moved. The experiment shows that this method can obtain the bandwidth of experimental data quickly and accurately, the Kullback-Leibler divergence between the real components and the false ones has clearly difference, and the false component can be accurately identified according to the threshold.

关 键 词：计量学 虚假分量 EMD K—L散度 遗传算法 窗宽 

分 类 号：TB973[一般工业技术—计量学]