Based On Higher-order Statistics Of The Magnetotelluric Impedance Estimates Of Applied Research

In signal processing,the traditional spectrum estimation is based on the theory of Fourier transform,and the conventional Fourier transform in signal processing put forward many condition,namely,signals or noise are assumed to be gaussian distribution,to fulfill the conditions based on second order statistics of Fourier transform can extract information,parameter identification and so on.For non-gaussian signal,it does not meet the basic assumptions of traditional spectrum estimation,so the traditional spectrum estimation will inevitably produce error,at the same time Fourier transform based on second order statistics does not include the phase information.Therefore,for non-gaussian and nonstationary and non-minimum phase signal,the second order statistics based on Fourier transform is powerless,and signal of higher order spectrum contains not only a lot of rich information that the second order statistics does not,but also can suppress gaussian or colored non-gaussian noise.Actually magnetotelluric signal with nonlinear and non-stationary,non-minimum phase characteristics,so the traditional power spectrum method cannot meet the requirement of the spectral analysis.By using conventional power spectrum to analyze the MT signals,the apparent resistivity and phase curves tend to be individual frequency dispersion,large error bar,abnormal shape,so the inversion processing can't get good geological interpretation results.To solve above problems,this paper put forward based on higher order statistics of the magnetotelluric signal spectrum estimation method,higher-order spectrum to reconstruct power spectrum of the signal,and then by using power spectrum estimation to recovery of the apparent resistivity and phase.First of all,this paper studies the generation of analog magnetotelluric signals under the condition of load noise,compared to traditional Fourier method and high-order cumulants method of the impedance of the estimated effect.Second for EH4 data measured,verify the magnetotelluric signal is non-gaussian,non-linear,non-minimum phase signal.Finally,based on high order system measurement and traditional Fourier method,complete the transformation from the original time series for the apparent resistivity and phase,and compare the two methods in the application of the actual data processing results.The results show that using higher order statistics method in magnetotelluric sounding signal spectrum estimation to suppress gaussian and non-gaussian colored noise,improve signal-to-noise ratio;Higher order statistics method eliminates the traditional power spectrum method of apparent resistivity and phase of individual frequency dispersion and form abnormal phenomenon,improves the geological parameters estimation precision and the interpretability of data.Therefore,compared with the traditional method of Fourier spectrum estimation,the higher-order cumulant spectrum estimation in magnetotelluric de-noising,spectrum analysis and impedance estimation has some advantages,for the magnetotelluric signal processing provides a new way.