中文摘要：

经验模态分解（EMD）算法非常适合非稳定序列信号、非线性序列信号以及复杂信号的分解，具有很高的噪声比。序列信号经过EMD分解为本征模函数（IMF）以及残差序列，所分解出来的IMF包含了原序列信号不同时间尺度的局部特征信号，是整个原序列的“去杂”反映。针对IMF所包含的不同尺度的特征这一特性，给出用EMD分解原始序列信号，提取其全部有限个本征模函数和残差序列，根据不同的IMF所包含原序列的特征信息量的大小引入信息权重训，然后通过欧氏距离对各个序列不同IMF序列进行相似匹配判定，最后通过综合各个IMF所占权重综合判定时间序列的相似匹配。实验结果表明，基于IMF对时间序列相似匹配和直接对原时间序列进行匹配，前者首先对时间序列进行分解，去掉其噪声等干扰，提取出IMF间接进行加权匹配，提高了时间序列的模式匹配精度，证明了该方法的有效性。

英文摘要：

Empirical mode decomposition （EMD） algorithm is very suitable for non-stable decomposition of the sequence of signals, nonlinear sequence signal, and complex signals with high noise ratio. Sequence signal after EMD decomposition the costs intrinsic mode functions （IMF） and the residual series decomposition IMF contains the local features of the different time scales of the original sequence signal is the original sequence to the complex reflects. Aiming at the IMF contained different scales characteristic of this feature, this paper gave EMD decomposition of the original sequence signal, extract all of its limit- ed number of intrinsic mode functions and residual series, the amount of information contained in the original sequence charac- teristics depending on the IMF the size of the introduction of the right to information, and then used the Euclidean distance to match the different IMF sequences for each sequence similarity, at last it comprehensively judged the time series similarity matching of each IMF＇s weight. The experimental results show that, time series similarity matching and directly matching to the original time series based on IMF, the former method firstly decomposed the time series to remove the noise and other interference, extracted IMF weighted matching indirectly improve the time series pattern matching accuracy, to prove the effectiveness of this method.