Application Of Time-frequency Analysis In Seismic Signal Processing

Time-frequency analysis has been used in many domain as a new signal processing method,especially in non-stationary signal.Seismic signal is a typical non-stationary.The application of time-frequency analysis is seismic spectrum characteristic,terrestrial magnetism,noise removing and energy compensation.With the development of mathematic method and the increasing of process capability,the profundity and scope of time-frequency analysis application is also broaden.The spectrum of seismic signal contains lithologic and structure information of stratum,so the spectrum analysis can interprate stratum lithological characters by looking for the internal relation and variation discipline.S transform and matching pursuit are used in seismic signal in the paper to predict thinbed and Direct Hydrocarbon Indicators.Short Fourier transform,WVD,wavelet transform,S transform and MPD are introduced first in the paper,and their relative merits are also discussed.S transform exerts large utility potential as its characteristic.Seismic response and spectrum characteristic of thin bed is analyzed.The Widess wedge model is used as example to discuss the influence results of the numbers of alternation of beds,reflector and its polarity.General S transform based on simplified windows is used to computate seismic signal spectrum,and according to the relation between tune depth and amplitude energy to predict thinbed by extracting single frequency profile.The process result of actual seismic data express the method can distinguish thinbed effectively.Usually the seismic wavelet is unknown,thus it is difficult to determine the master frequency by time domain and the frequency peak only can be determined by spectrum.At the same time,the relation between master frequency and peak frequency is uncertain,so it is rational that the peak frequency is usually regarded as master frequency.Combined with the master parameter computation of complex track analysis,master frequency analysis based on general S transform is adopted to inscribe the bottom and top of thinbed and have better process effect.The lithological characters can be determined by obtaining the information such as amplitude,frequency and formation velocity,which can be used to look for oil and gas directly.Low frequency shadow is that the strong energy instanous low frequency zone under the oil and gas reservoir,which is one of the important attribute of reservoir.As wavelet transform parameters influence directly the time-frequency resolution of the result.The few numbers of sample in scale domain make the transform resolution decreasing and mosaic phenomenon.Thus,General S transform based on adjustable complex windows is adopted to determine low frequency shaow.The processing results demonstrate that the method has practical worthiness and provide a new think for oil and gas indicator.The common orthogonal decomposition such as Fourier transform,short Fourier transform and wavelet transform don't consider the self characteristic of signal because the orthogonal basis augmentation is to use an finite funcation or vector assemble to express arbitrary signal.In addition,the express ability and scope of signal are also limited because of the finite basis funcation.The think of split decomposition can select basis funcation adaptively according to the signal characteristic,which need fewer funcations to express tha basic characteristic.MP algorithm is a kind of signa decomposition method based on over-complete basis,which select suitable dictionary to match original signal and obtain the spit expression of signal.MP algorithm is used to seismic signal process,Particle Swarm Optimization and simple type local optimization is combined to optimize the atom because of the big computation of MP algorithm.The results demonstrate that computation is decreased effectively.Noise removing and reconstructure,DHI are acted as example to validate the validity of the method.At last,the work and harvest is summarized,the deficiency and future direction are also indicated.