基于夹角余弦和模糊阈值的EEMD去噪方法

针对非线性非平稳信号的去噪问题，结合EEMD分解信号的自适应特性，提出一种基于夹角余弦和模糊阈值的去噪方法。首先用夹角余弦法计算各个本征模态函数（IMF）与观测信号之间的相似度，以相似度曲线的首个极小值的后一个位置为分界点将分解出的IMF分为噪声主导模态和信号主导模态；然后根据VisuShrink阈值易“过扼杀”细节系数和SUREShrink阈值易“过保留”噪声系数的特点，利用模糊阈值对噪声主导的IMF进行处理；最后将所有的IMF重构得到消噪信号。分别采用仿真信号和真实ECG信号进行去噪实验。结果表明，所提方法在整体性能上优于小波半软阈值方法、基于EMD的软阈值（EMD-Soft）和间隔阈值（EMD-IT）方法，是一种有效的去噪方法。

A noise suppression scheme with EEMD based on angle cosine and fuzzy threshold

In this paper,a novel noise suppression scheme that exploits the angle cosine and fuzzy threshold,is pro?posed to improve the denoising accuracy for nonlinear and non-stationary signals. The proposed scheme combines with the adaptive characteristics of the conventional EEMD as well. Firstly,the similarity between intrinsic mode function(IMF)and observed signal is calculated by the angle cosine method. The IMFs are then divided into noise dominated and signal dominated part by finding the latter location of the first minimal value of cosine similarity curve. The VisuShrink threshold tends to“overkill”the detail coefficient,while the SUREShrink threshold tends to“underkill”the noise coefficient. Employing a fuzzy threshold that is an optimal value within the convergent range of VisuShrink and SUREShrink threshold,we can remove the noise part from the noise dominated IMF well. Finally,all of the remained IMFs are reconstructed to get a noise suppressed signal. Simulation experiments are conducted by us?ing both suppositional and actual ECG signal. The simulation results demonstrate that the proposed scheme can effec?tively suppress the noise,and has better denoising performance than some conventional schemes,such as the half soft threshold wavelet method,EMD soft threshold(EMD-soft)method and the EMD interval threshold(EMD-IT)method.