Local Integral Mean-Based Sifting for Empirical Mode Decomposition

A novel sifting method based on the concept of local integral mean of a signal is developed for empirical mode decomposition (EMD), aiming at decomposing those modes whose frequencies are within an octave. Instead of averaging the upper and lower envelopes, the proposed technique computes the local mean curve of a signal by interpolating data points that are local integral averages over segments between successive extrema of the signal. With the sifting method, EMD can separate intrinsic modes of oscillations with frequency ratios up to 0.8, thus considerably improving the frequency resolving power. Also, it is shown that the integral property of the sifting considerably accelerates the convergence of the sifting iteration and remarkably enhances the robustness of EMD against noise disturbance.