The Application Of The Hilbert-Huang Transform In Seismic Data Processing

HHT has some advantages in nonlinear and nonstationary signal analysis over traditionaltime-frequency analysis method. Data can be decomposed in terms of IMFs according to itscharacteristics by HHT and we don’t need to consider other interferences,such as the choiceof basis function and window function. Therefore, the result can reflect the characteristics ofthe signal itself. The core of HHT is Empirical Mode Decomposition, which has merits inprocessing signals. However, there are some serious drawback of it and one of them is themode mixing problem. To deal with the mode mixing problem, we discuss some improvedalgorithms,including Ensemble EMD (EEMD), Complementary Ensemble EMD and EMDwith average value constraint. EEMD can provide a uniformly distributed reference scale andthe extreme point features can also be changed, which essentially resolve the mode mixingproblem associated with EMD through the help of added noises. CEEMD makeimprovements based on the EEMD In the new CEEMD, white noise is added in pairs to theoriginal data and the paired noises could eliminate the residue of added white noises totally.The computation speed has been improved either. The EMD with average value constraintconsiders the small change of the strong low frequency signal when we calculate the envelope,which can reflect the change of the characteristics of low-order and weak mutation point well.In aspect of denoising, this paper mainly adopts the frequency separating filter methodand wavelet threshold method based on CEEMD. The frequency separating filter methoddenoise by abandoning components that contain more noise. Wavelet threshold method basedon CEEMD has great advantages over the CEEMD frequency filter and wavelet thresholddenoising and has well application prospect. F-x domain EMD and EMD frequencyseparating filter method are given in this paper to remove the surface wave. F-x domain EMDutilizes the characteristics of low frequency and high wave number of the surface wave toremove it. EMD frequency separating filter method mainly considers the frequency domaincharacteristics of surface wave and the orthogonal hypothesis of surface wave and primaryreflections. EMD decomposition improves the resolution of seismic section in view of thedeep reflection wave of poststack data, and the frequency band is broaden properly. Inaddition, we attempt to denoise by2D empirical mode decomposition based on EEMD andobtain a certain effect.In general, modeling computation and practical data processing prove that the effect ofimproved EMD is better than that of EMD. HHT has achieved satisfactory results in theapplication of denoising, resolution improvement and so on.