A New Wavelet-Based Mode Decomposition for Oscillating Signals and Comparison with the Empirical Mode Decomposition

摘要

We introduce a new method based on wavelets (EWMD)for decomposing a signal into quasi-periodic oscillating components withsmooth time-varying amplitudes. This method is inspired by both the“classic” wavelet-based decomposition and the empirical mode decomposition(EMD). We compare the reconstruction skills and the perioddetection ability of the method with the well-established EMD on toysexamples and the ENSO climate index. It appears that the EWMD accuratelydecomposes and reconstructs a given signal (with the same efficiencyas the EMD), it is better at detecting prescribed periods and isless sensitive to noise. This work provides the first version of the EWMD.Even though there is still room for improvement, it turns out that preliminaryresults are highly promising.