基于灰色预测模型的EMD端点效应抑制方法  

A Method to Suppress the End Effect of EMD Based on GM(1,1)Prediction Model

作　　者：钱欢 蒋沅[1] 王海涛 QIAN Huan;JIANG Yuan;WANG Hai-Tao(School of Information Architectural Engineering,Nanchang Hangkong University,Nanchang 330063,China;The Second Construction Co.,Ltd.of China Construction First Group,Beijing 102600,China)

机构地区：[1]南昌航空大学信息工程学院,南昌330063 [2]中建一局集团第二建筑有限公司,北京102600

出　　处：《南昌航空大学学报（自然科学版）》2021年第2期21-26,共6页Journal of Nanchang Hangkong University(Natural Sciences)

基　　金：国家自然科学基金(61663030,61663032);江西省研究生教改项目(JXYJG-2017-131)。

摘　　要：为了能够有效地抑制经验模态分解(EMD)中存在的端点效应问题,本文提出一种基于数学模型和信号分析法相结合的处理方法——基于灰色预测模型的EMD端点效应抑制方法。首先利用灰色预测模型对原始信号的两端进行延拓产生有限极值点,对延拓信号的中间部分加矩形窗保证原信号不变,对延拓部分加汉宁窗让包络线拟合更加平滑,然后对处理过的信号序列进行EMD分解得到各阶固有模态分量(IMF),最后舍弃掉IMF中的延拓部分,从而实现对端点效应的有效抑制。通过对仿真信号和实验信号进行处理分析,以均方根误差法和相似系数法作为量化评价指标,将处理前后的IMF分量和原信号的组成分量作比较,表明本文改进的方法能够实现抑制端点效应的目的。In order to effectively suppress the end effect in empirical mode decomposition(EMD),this paper proposes a processing method based on the combination of mathematical model and signal analysis method which the end effect suppression method of EMD based on grey prediction model.Firstly,the gray prediction model is used to extend the two ends of the original signal to produce limited extreme points.A rectangular window is added to the middle part of the extended signal to keep the original signal unchanged,and a Hanning window is added to the extended part to make the envelope fitting smoother.Then EMD is used to decompose the processed signal sequence to obtain the intrinsic mode components(IMF).Finally,the extended part of the IMF is discarded,so as to reduce the noise It can effectively suppress the endpoint effect.By processing and analyzing the simulation signal and the experimental signal,taking the root mean square error method and the similarity coefficient method as the quantitative evaluation index,the IMF component before and after processing is compared with the component of the original signal,which shows that the improved method in this paper can achieve the purpose of suppressing the endpoint effect.

关 键 词：HHT 经验模态分解 端点效应 灰色预测模型 窗函数 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]