Adaptive Time-Frequency Peak Filtering And Its Application On Noise Attenuation For Seismic Data

With the rapid growth of the petroleum industrial, what is most wanted is the advanced theory of the gas prospecting and the technique that can solve difficult problems in production. Seismic exploration is an important tool for oil-gas reservoirs and mineral resources exploration. With the development of method of seismicexploration and increasing demand for oil-gas, the prospecting has turn to research the complex structure and deep crustal structure. The seismic exploration data obtained from the complex earth's surface, such as desert, hill and boondocks etc., is rather complex and has strong random noise. Attenuation of noise and improvement of SNR is significant for researching the structure of subsurface and for prospecting oil-gas reservoirs.IN practice, signals are often corrupted by noise, and this has the effect of hindering the recovery of important information encoded in the signal. All signal processing fields including radar, sonar, communications, seismology, and biomedicine suffer from this problem. The performance of signal enhancement algorithms generally depends on the signal-to-noise ratio (SNR). Many signal processing algorithms work well for high SNR situations, but most perform poorly when SNR decreases below a given threshold. In this case, if no adequate alternative algorithm is available, preprocessing or filtering is required to improve the SNR.Adaptive filters perform poorly in certain conditions, however. An example of this is filtering of a nonstationary signal whose spectral content changes quickly with time. The filter designed using LMS approach, for instance, may not adapt quickly enough to track the rapidly changing signal. This is due to the delayed convergence of the algorithm, which is a function of the signal autocorrelation matrix eigenvalues. Further, adaptive methods require that the structure of the filter (such as the number of the taps) and an estimate of SNR be imposed for optimal performance. This may not be possible or, in some situations, advisable. In many signal processing applications, the structure of the underlying signal is often unknown and too complicated to model accurately. Any attempt to provide a model framework will lead to suboptimal results and even to erroneous conclusions about the signal.Time-frequency peak filtering(TFPF) is a novel algorithm for signal enhancement based on Time-frequency analysis which can remove the noise effectively and reconstruct the original signal. It allows the reconstruction of signals from observations corrupted by additive noise by encoding the noisy signal as the instantaneous frequency (IF) of a frequency modulated(FM) analytic signal. IF estimation is then performed on the analytic signal using the peak of a time-frequency distribution(TFD) to recover the filtered signal. For deterministic band-limited non-stationary multi-component signals in additive white Gaussian noise (WGN), the IF estimation using the pseudo Wigner-Ville distribution(PWVD) is approximately unbiased. Testing of the method on synthetic mono-component and multi-component signals shows clean recovery of the signals in noise level down to a signal-to-noise ratio(SNR) of -9 dB.In this paper, based on the discrete algorithm for Time-frequency peak filtering(DTFPF), we propose a modified algorithm named adaptive time-frequency peak filtering(ATFPF) which improves the accuracy of IF estimation. If the IF is a nonlinear function of time, the bias of the estimate depends on the window length. In ATFPF, based on the asymptotic formulae for the variance and bias, we modify the method of IF estimation based on PWVD and build a new IF estimator with a time-varying and data-driven window length. This new technique provides a new choice for people to resolve the bias-variance tradeoff. In the last part, we test ATFPF on piecewise constant IF with a step, piecewise linear frequency modulated signal with a step and a nonlinear frequency modulated signal respectively, and show the simulation results comparing with the original TFPF with time-invariant window lengths. Additionally, the normalized mean absolute errors are calculated for different window lengths providing a further comparison. From the contrast of the original and the proposed adaptive algorithms, it can be concluded that the second one shows a much better accuracy especially for signals with a step and nonlinear signals.Seismic data is the important information resource of geological prospect and exploration. Random noise in seismic reflection data can be introduced by various sources and is often a problem in geophysical data visualization because it obscures fine details and complicates identification of image features.According to the rule of from simple to complex condition, the paper firstly applied ATFPF to the synthetic seismic data, then to actual seismic data. The constructed seismic data with single event considered the different factors of seismic wave, including the velocity of event, the frequency of seismic wave, and the attenuation of frequency with in an event, phase-to-amplitude radio of seismic wavelet as well as the mixed phased wavelet. The white random noise is added into the above models to simulate TFPF, the results show that the random noise in the seismic data with low SNR are attenuated validly, the reflective signal is enhanced and SNR of seismic data is increased .The wave of recovered signal is similar to the true wave form with the correct position at wave peak and wave valley.Compared with the result of the TFPF,the ATFPF can suppress the noise better,and also restore the original signal better. Because ATFPF is flexible tool for attenuating added random noise, In this paper, ATFPF were applied to the actual shot of the seismic exploration signals, reduction From the results we can see that the strong noise concentrated in some area is attenuated, the reflected events are enhanced and the continuity of event is improved. Profile data processed by ATFPF shows more events in filtered data than those in data before ATFPF processing for the attenuation of random noise and enhancement of signal,recover the effective seismic information very well.Through theoretical analysis, as well as the actual seismic signal processing, we can see that the method can reduce the random noise of the seismic data and the restoration of useful information effectively. Applicable to the unknown spectrum and non-stationary signals. Thus, it is feasible to use the ATFPF algorithm to reduce the random noise in seismic data, it is valuable not only in the theoretical research, but also in the actual seismic exploration as well as other applications that need random noise reduction,and it will have good applications prospects.