MT时间序列的小波去噪分析

WAVELET-BASED DENOISING OF MT TIME SERIES

从本质上说 ,MT时间序列中噪声的强度与类型是能否取得MT响应参数无偏估计的决定性因素。当MT时间序列中磁场和电场中都含有相关噪声时 ,传统的去噪方法已无能为力。结合小波分析与MT时间序列的特征 ,提出了一种基于小波分析的MT时间序列去噪方法 ,讨论了基于小波分析的噪声识别 ,分析了理论数据通过小波分解与重构实现的去噪处理 ,探讨了对实测时间序列的固定源和随机干扰的去噪处理。

Acting as a major method of geophysical prospecting, Magnetotelluric Sounding (MT) is effective in oil and gas prospecting, geothermal survey and deeper earth prospecting. But the undeveloped data processing leads that MT has low resolution, which is the major factor that retards MT from being widely used. There is no method being able to obtain unbiased estimates of the transfer function when there is stonger correlated noise in both the electric and the magnetic time series. Wavelet analysis could decompose the composite signal, which consists of several components of different frequencies, into a series of signal block. So it is an effective method for dissociating the noise from signal. In this paper, we use wavelet analysis to denoise MT data by decomposition and reconstruction. The result is more acceptable than the conventional denoised result.