Modified signal de-noising approach for multiplicative noise based on empirical mode decomposition

Multiplicative noise is usually caused by non-ideal(non-stationary or nonlinear) channel. The signal is multiplied by the noise in mixture model. Therefore, it is difficult to remove it. Under the background of multiplicative noise reduction, multiplicative noise can be converted into additive one, by introducing the well-known homomorphic transformation to remove the coupling between noise and signal. Characteristics of energy distribution of harmonic signal and its homomorphic version, disturbed by uniform white noise, are further studied using empirical mode decomposition(EMD), which leads to one new amplitude-thresholding rule of intrinsic mode functions(IMFs), adaptive to the application requirement of multiplicative noise reduction. Thus, modified EMD based signal de-noising approaches for multiplicative noise reduction are proposed. Experimental results show that the best de-noising performance is given by the modified algorithm with soft threshold, i. e. HEMDA-S. Furthermore, high reconstruction precision of source signal can be already assured by the modified de-noising algorithm, by using the constructed amplitude-thresholding rule of IMFs based on second-order polynomial regression analysis. The mismatch between the used amplitude-threshold on IMFs and their actual noise levels may occur if the order of used regression polynomial is too high, which will significantly reduce de-noising performance of the modified algorithm. © 2015 Journal of Mechanical Engineering.