EMD was proposed by Bekara and van der Baan (2009) to attenuate random noise. They applied EMD on
each frequency slice in the
domain, and remove the first IMF,
which mainly represent the higher wavenumber components. The methodology can be summarized as:
denote the estimated signal and acquired noisy signal, respectively.
denote the forward and inverse Fourier transforms along the time axis, respectively.
th EMD decomposed component.
denotes frequency. However, a problem occurs when applying
EMD, because the dipping events will also be removed.
This problem occurs because, for many data sets, the random noise and any steeply dipping coherent
energy make a significantly larger contribution to the high-wavenumber energy in the
domain than any desired signal (Bekara and van der Baan, 2009).
Random noise attenuation by a selective hybrid approach using f-x empirical mode decomposition