EMD interval thresholding denoising based on similarity measure to select relevant modes

This paper introduces a novel EMD interval thresholding (EMD-IT) denoising, where relevant modes are selected using a l₂-norm measure between the probability density function (pdf) of the input and that of each mode, thresholds are estimated by the characteristics of fractional Gaussian noise (fGn) through EMD. To solve the problem of more relevant modes included when the signal is corrupted by fGn with the H increase, a modified l₂-norm method was given. The computational complexity of EMD-IT denoising is also analyzed. And the time complexity of it is equal to that of EMD. Numerical simulation and real data test were carried out to evaluate the effectiveness of the proposed method. Other traditional denoisings, such as correlation-based EMD partial reconstruction (EMD-PR), EMD direct thresholding (EMD-DT) and NeighCoeff-db4 wavelet denoising are investigated to provide a comparison with the proposed one. Simulation and test results show its superior performance over other traditional denoisings in whole.