Strong noise separation for magnetotelluric data based on a signal reconstruction algorithm of compressive sensing

Tang Jing-Tian,Li Guang,Xiao Xiao,Li Jin,Zhou Cong,Zhu Hui-Jie

DOI: https://doi.org/10.6038/cjg20170928

2017-01-01

Abstract:To suppress strong noise in raw magnetotelluric (MT) data, we propose a new time series denoising method based on a signal reconstruction algorithm of compressive sensing. A redundant dictionary which matches with strong noise but insensitive to useful MT signal is designed, then the strong noise is separated from raw MT. data by using this dictionary and an improved orthogonal matching pursuit algorithm. In order to verify the effect of the proposed method, firstly, a strong noise separation simulation experiment is carried out by adding ideal strong noise into the measured MT signal. Secondly, three time domain data segments with different types of typical strong noise selected from a large number of measured data are used for strong noise separation test. Finally, the proposed method is applied to measured data collected in the Qaidam Basin and Lu-Zong ore-concentration area. Simulation results show that the proposed method can effectively separate the strong noise from raw data while keep the useful part. The results of the applications to measured data show that our proposed method is an effective method to suppress strong noise and therefore improve the quality of MT data collected in areas with strong interference.