Seismic data time-frequency analysis by the improved Hilbert-Huang transform

As deficiencies of the conventional Hilbert-Huang transform(HHT),the oscillation mode mixing interference in the intrinsic mode function(IMF) components is generated by the conventional empirical mode decomposition(EMD).In order to eliminate this mode,this paper proposes an improved Hilbert-Huang transform method.Firstly,the wavelet packet transform(WPT) is used to decompose a signal into a few narrow band signals.Then EMD decompose these narrow band signals into intrinsic mode functions(IMFs).After that,the true IMFs are kept and unrelated IMFs are removed from the result according to the correlation coefficient method.Finally Hilbert transform instantaneous frequency and improved HHT amplitude are obtained.The improved HHT method gets rid of the Heisenberg uncertainty principle limits of wavelet transform,and the oscillation mode mixing interference of the conventional HHT.Although improved HHT amplitude have jump phenomenon in high frequency parts,they are all around the real frequency value and have narrow frequency band.This new method can reflect the real characteristics of frequency.The numerical and real data examples show that the improved method has feasibility in seismic data analysis.