为了改善脉冲星辐射脉冲信号的消噪效果，提出了一种基于噪声模态单元预判的经验模态分解(EMD)消噪声方法。该方法首先利用EMD将含噪辐射脉冲信号分解为一组内蕴模态函数(IMF)，根据IMF系数的统计特性采用局部均方误差准则进行噪声模态单元预判，并将噪声模态单元置零；然后对噪声模态单元预判处理后的IMF以模态单元为基本单位进行最优比例萎缩消噪，从而达到抑制噪声、保留信号的目的。实验结果表明：与Sure Shrink小波阈值法、Bayes Shrink小波阈值法和EMD模态单元比例萎缩法相比，基于噪声模态单元预判的EMD消噪方法可以更有效地去除脉冲辐射信号中的噪声，同时更好地保留信号突变处的细节信息特征，在信噪比、均方误差、峰值相对误差、峰位误差和相位误差等方面都有一定程度的改善%In order to improve the de-noising effect of the pulsar signal, an empirical mode decomposition (EMD) denoising algorithm based on the prediction of noise mode cell is put forward. The core steps of the proposed method is as follows: firstly, the noisy pulsar signal is decomposed into a group intrinsic mode function (IMF) by EMD, and the noise mode cell is predicted according to the IMF coefficients statistics and local minimum mean square error criteria. The selected noise mode cells are set to be zero. Then the IMF which has been processed according to noise mode cell prediction is denoised by optimal mode cell proportion shrinking, for removing the noise and retaining the signal details. The experimental results show that compared with the Sure Shrink wavelet threshold algorithm, Bayes Shrink wavelet threshold algorithm and the EMD mode cell proportion shrinking algorithm, the proposed method performs well in removing the pulsar signal noise and retaining the signal details information. The proposed method can achieve a higher signal-to-noise, the lower root mean square error, error of the peak position, relative error of the peak value and phase error.
展开▼
