摘要
经验模态分解(EMD)是针对非线性和非平稳数据的有效分析方法,但是原始算法有多余分量、分量之间不完全正交等缺点.本文引入筛选系数λ将原始EMD算法推广为广义EMD算法,并且使用最小化正交条件来选取最优筛选系数.模拟数据和实际数据的分析结果显示,相比于原始EMD算法,该算法有效地减少了多余分量,更好地分解出了时间序列的趋势成分,而且提高了IMF成分序列之间的正交性.由于筛选系数是数据本身决定的,因此该算法比原始算法有更强的自适应性.
Empirical mode decomposition is an efficient method for non-linear and non-stationary data. But original algorithm always produces redundant components, and orthogonaltiy is not ensured. This paper modifies the original EMD algorithm by choosing the optimal sifting parameter based on the orthogonal criterion in every sifting process. The numerical experiments show that, comparing with original algorithm, modified algorithm can reduce the redundant components efficiently, determine the trend component more accurately and improve the orthogonality between IMF components. As the sifting parameter is determined adaptively by data itself, the modified algorithm is more adaptive than original algorithm.
出处
《数理统计与管理》
CSSCI
北大核心
2013年第6期1002-1012,共11页
Journal of Applied Statistics and Management
