Signal-Noise Identification of Magnetotelluric Signals Using Fractal-Entropy and Clustering Algorithm for Targeted De-Noising

Abstract

A new technique is proposed for signal-noise identification and targeted de-noising of Magnetotelluric (MT) signals. This method is based on fractal-entropy and clustering algorithm, which automatically identifies signal sections corrupted by common interference (square, triangle and pulse waves), enabling targeted de-noising and preventing the loss of useful information in filtering. To implement the technique, four characteristic parameters — fractal box dimension (FBD), higuchi fractal dimension (HFD), fuzzy entropy (FuEn) and approximate entropy (ApEn) — are extracted from MT time-series. The fuzzy c-means (FCM) clustering technique is used to analyze the characteristic parameters and automatically distinguish signals with strong interference from the rest. The wavelet threshold (WT) de-noising method is used only to suppress the identified strong interference in selected signal sections. The technique is validated through signal samples with known interference, before being applied to a set of field measured MT/Audio Magnetotelluric (AMT) data. Compared with the conventional de-noising strategy that blindly applies the filter to the overall dataset, the proposed method can automatically identify and purposefully suppress the intermittent interference in the MT/AMT signal. The resulted apparent resistivity-phase curve is more continuous and smooth, and the slow-change trend in the low-frequency range is more precisely reserved. Moreover, the characteristic of the target-filtered MT/AMT signal is close to the essential characteristic of the natural field, and the result more accurately reflects the inherent electrical structure information of the measured site.

Separation of Magnetotelluric Signals Based on Refined Composite Multiscale Dispersion Entropy and Orthogonal Matching Pursuit

“…According to the differences of various entropies in the sample library signal which include the 150 data samples that are from MT field measurements, refer as (Li et al 2018). Among them, 50 data samples without interference are from the unmanned areas, and the remaining 100 data samples are collected from the strong electromagnetic interference areas.…”

“…Li J et al combined VMD with detrended fluctuation analysis (DFA) to adaptively select K value and the number of the reconstructed modals, thus improving the de-noising effect . In addition, the original MT signal-noise identification method is based on fractal, entropy, and complexity, and multi-feature cluster analysis, and the fidelity of the data is improved (Li et al 2018). By removing noises in time domain, the statistical analysis and time-series editing methods can directly and effectively improve the quality of MT data, but they could also damage the effective slow portion on the noisy segments.…”

“…The proposed method is compared with the remote reference method and the OMP-based overall method, which can purposefully remove the identified noise and retain the low-frequency useful MT signal. Compared with the fractal-entropy and clustering method (Li et al 2018) in the analysis of multiple characteristics of MT data, it makes more precise and efficient features extraction of MT data that improves the efficiency of MT signal-noise identification. Moreover, the experimental results at the measured sites of apparent resistivity-phase curves and polarization direction reflect that the proposed method can more reliably approach to the natural MT field, and truly reflect the underground electrical structural information.…”

Audio Magnetotelluric Signal-Noise Identification and Separation Based on Multifractal Spectrum and Matching Pursuit

“…The sample library contains four types of typical MT data. 24 One of these is MT data basically without interference, collected in regions where population is sparse and there is no strong electromagnetic interference. The characteristics of MT data without interference are close those of natural electromagnetic fields.…”

An Introduction to Fractal-Based Approaches in Unconventional Reservoirs — Part I

“…Magnetotelluric was an electromagnetic-based exploration method, which has been widely used to identify the distribution of underground geoelectric structure. Jin Li et al 23 proposed a new technique for signal-noise identification and targeted de-noising of Magnetotelluric signals based on fractal-entropy and clustering algorithm.…”