Noise assisted signal decomposition method based on complex empirical mode decomposition

Abstract

The ensemble empirical mode decomposition has been proposed in order to alleviate mode mixing in empirical mode decomposition, but the ensemble average in it can always result in new mode mixing, spectrum losing, and computational cost increasing, which can affect the analysis and extraction of signal physical characteristics. To tackle these problems, a noise-assisted signal decomposition method based on complex empirical mode decomposition is proposed, in which the mode mixing is reduced by taking the projection of intrinsic mode functions decomposed from white noise as basis functions for signal extrema extraction. While the problems result from ensemble average are reduced because the effects of noise projection are eliminated in the process of calculating the envelope barycenter. Simulation results show that our method has further reduced mode mixing, and speeded up the operation rate visibly and alleviated spectrum losing to a certain degree.