作者投稿和查稿 主编审稿 专家审稿 编委审稿 远程编辑

计算机工程

• 移动互联与通信技术 • 上一篇    下一篇

多小波包下经验模态分解去噪研究

王唯嘉 1,肖明清 1,张磊 1,陈茂才 2   

  1. (1.空军工程大学航空航天工程学院,西安 710038;2.中国船舶重工集团公司第704研究所,上海 200031)
  • 收稿日期:2015-07-06 出版日期:2015-12-15 发布日期:2015-12-15
  • 作者简介:王唯嘉(1990-),男,硕士研究生,主研方向:信号处理与去噪;肖明清,教授、博士;张磊,博士研究生;陈茂才,工程师、硕士。

Research on Empirical Mode Decomposition Denoising Under Multiwavelet Packet

WANG Weijia 1,XIAO Mingqing 1,ZHANG Lei 1,CHEN Maocai 2   

  1. (1.School of Aeronautics and Astronautics Engineering,Air Force Engineering University,Xi’an 710038,China; 2.The 704th Research Institution,China Shipbuilding Industry Corporation,Shanghai 200031,China)
  • Received:2015-07-06 Online:2015-12-15 Published:2015-12-15

摘要:

传统基于小波分析的去噪方法和经验模态分解(EMD)去噪方法去噪后的信号信噪比较低。针对该问题,提出一种多小波包框架下区间迭代不变阈值的EMD去噪方法。对输入带噪信号进行预处理,将其变换为多维信号之后进行多小波包分解,设计针对软硬阈值函数的改进型阈值函数,并对得到的最后一层多小波包系数实现小波阈值处理,从而得到一维小波系数,对各本征模态函数分量(IMF)进行区间迭代不变阈值EMD去噪,并重构得到去噪后信号。仿真结果表明,与传统EMD小波阈值去噪方法相比,该方法信噪比提升近2.5 dB,均方误差达到0.000 7, 去噪效果较好。

关键词: 多小波包, 经验模态分解, 函数拟合, 新阈值函数, 阈值去噪, 平移不变量

Abstract:

The Signal to Noise Ratio(SNR) of the denoised signal is low by using the approaches of traditional denoising method based on wavelet analysis and the method based on Empirical Mode Decomposition(EMD).Aiming at this problem,this paper proposes an interval iterative threshold of EMD denoising method under the frame of multiwavelet packets.The input signal with noise is preprocessed firstly.After the signal is transformed into the multidimensional form,the paper devises a new thresholding function which is inspired by the hard and the soft thresholding function.The multi-wavelet coefficients in the last level is denoised by the new thresholding function and the one dimensional wavelet coefficients is gotten,which is followed by the Intrinsic Mode Function(IMF) deposed by this method.The denoised signal is acquired by reconstruction.Simulation results show that the novel method can improve the SNR by about 2.5 dB and the Mean Squared Error(MSE) also decreases to 0.000 7 compared with the traditional methods,and the ability of denoising is better.

Key words: multiwavelet packet, Empirical Mode Decomposition(EMD), function fitting, new threshold function, threshold denoising, translation invariant

中图分类号: