Research On MRS Signal Denoising Method Based On Matching Pursuit

As everyone knows,but Chinese although vast territory and abundant resources face a series of problems increasingly serious environmental and population,fresh water resource is very scarce,it also forced the groundwater exploration and utilization has become the research focus in the present stage.Magnetic resonance sounding(MRS)technology is the most direct and effective geophysical detection method that can directly acquire groundwater information at present.It is widely used in water resources exploration,evaluation and dam leakage,and mine/tunnel water gushing.With the continuous development of science and technology,although the MRS detection accuracy and detection instrument in depth has been improved,but because the MRS signal is nanovolt level very weak signal,the signal collected by environment easily in the complex spike noise and random noise and harmonics noise interference,of which the most serious is the random noise and the frequency of harmonic noise.Strong noise interference reduces the accuracy of the characteristic parameters of the MRS signal,and then affects the subsequent inversion interpretation.Therefore,how to filter environmental noise and extract the characteristic parameters of MRS signals is a key issue in the field of groundwater detection.It is also one of the hotspots and difficulties in recent years in the field of NMR denoising.Matching pursuit(Matching Pursuit,MP)method as a tool of time-frequency analysis,through sparse decomposition in over complete dictionary will find the signal reconstruction and the matching of atoms,then according to the characteristics of noise eliminating noise of atoms,atoms are selected useful constructs,extract the useful signal.Because this method has the advantage of fine description of local information and so on,it has been widely applied in signal processing,such as signal feature extraction,denoising and reconstruction.Based on a large number of literature,this paper proposes a method of MRS signal denoising based on matching pursuit.The expected denoising accuracy is ?E0 within ±5%,AT2*within ±100%and the signal-to-noise ratio is between 15?25dB.First,according to the characteristics of the MRS signal,the common atom Library of the MP method is selected:Gabor and discrete cosine transform(DCT).By comparing the reconstruction results of two kinds of atomic libraries,we choose a better Gabor library with sparse decomposition effect.The MP denoising algorithm based on Gabor atomic library can meet the expected requirements,and the signal-to-noise ratio can reach 38dB.In order to verify the performance of the proposed algorithm,carried out with independent component analysis(Independent Component Analysis,ICA)contrast experiment method,The de noising performance of matching pursuit method based on Gabor atom library is verified,but its application scope is limited,and it has large computation complexity and high complexity when constructing over complete atomic library,which results in the long time and low accuracy of reconstruction process.Then,in view of the characteristics of the magnetic resonance signal itself and its frequency harmonic noise,a matching tracking algorithm based on the structural oscillating atom library is proposed.According to the attenuation coefficient P value denoising scheme put forward two steps:(1)defined p =0(this time for the harmonic oscillation),find a most optimal atom,frequency,frequency and the reconstruction of the lost,remove all power and so on until now;(2)the first step to remove frequency interference,?1/T2*(the residual signal contains only MRS signal and random noise),find the best matching atom and MRS signal directly.The reconstruction of signal extraction to eliminate random noise.The MP de-noising algorithm based on the oscillating atomic library can meet the expected requirements.Moreover,the error caused by the fluctuation of the frequency base frequency in the 50 ±0.2Hz range can be avoided by reasonably setting the search range and the step length of the p value.Finally,simulation and comparison with the most effective harmonic modeling method to filter power frequency interference further illustrate the universality of MP denoising algorithm based on oscillating atomic library.Finally,in view of the noise elimination ability of the traditional atom library and the oscillating atom library,the experimental data processing experiments are carried out.Using Gabor and oscillatory atom library MP denoising algorithm for processing groups based on the measured data,verify the feasibility and practicability of the proposed MRS signal noise filtering method of the matching pursuit.