Research On Methods And Its Application Of Separating The Signal From Noise Based On The Modern Spectrum Estimation

In many projects of signal process, the power spectrum can provide the power's distribution with the frequency, so it is very important. The power spectrum include nonparametric methods and parametric methods. In this paper, the parametric methods are mainly introduced to adopt forecast recursion measures, which are based on statistical. The purpose of the thesis is to introduce the modern spectral estimation into seismic data processing to separate the seismogram of the signal and noise, whose spectrum overlapped together, and improve the seismogram's signal-to-noise ratio (SNR), and we can use modern spectrum estimation to get higher resolution, we also can retrieve useful lithologic information, which provide validity and dependability for seismic exploration.Firstly, the academic introduced the basic thoery of modern spectral estimation, and discussed several main spectral estimation methods. Owing to the autoregressive (AR) model spectral estimation is the most common method in this thesis, and then dissertated the character and the choice of the prediction error rank, systemically summarized several important algorithm of estimating the autoregressive parameters, and analysed the deficiencies and advantages of the several algorithms are analyzed. Then this thesis aimed at the shortages of Burg algorithm and put forward a modified Burg algorithm, which restrained the disadvantages of Burg algorithm. Finally this thesis introduced some examples of signal noise ratio estimation, through estimating the SNR of seismogram, and then utilizing the AR spectral estimation estimated the noise's spectrum, eliminated the noise from seismic data, and improved the SNR of seismogram. The results of emulation experiments and the seismic data processing proved that the modern spectral estimations could present higher SNR, and the results almost didn't have distortion.