| abstracts | An adjustment about the added white noise in Ensemble Empirical Mode Decomposition (EEMD) of Wu and Huang (2008) was presented in this paper. The EEMD establishes an ensemble by adding time series of finite but not infinitesimal amplitude white noise into a time series of the signal to solve the mode-mixing problem occurred in the conventional EMD method. The adding ensembles of noise are supposed to be exhausted from all possible solutions from the sifting process. However, in the Matlab script of the theory, it was found that the added noise could not be averaged out through the whole process. The residue added noise thus causes some extra signals exist in the sifting results, even the number of trials of the ensemble was up to 5000. In the adjusted method, the added noises were randomly selected without repetition from a noise set which satisfies the normal distribution at each sampled node. With this approach, the added noises can be entirely averaged out without any residue at each node and on the entire time series. The experiments show that number of trials can be reduced to 50 sets. It not only avoid the time consuming problem but also retains the benefit of EEMD. |