Research Of Denoising Methods For MRS Signal Based On EMD

Recently, Magnetic Resonance Sounding (MRS) technology in detectingunderground water has got rapid development. Comparing with traditionalgeophysical exploration technology, this technology has many advantages, such ashigh resolution, high efficiency, low cost and abundant amount of information, etc. Asa result, MRS technology is the only geophysical method to detect underground waterdirectly. Essentially, this method makes energy level transition of hydrogen proton inunderground water. When the extranuclear electrons drop to a low level, the coinreceives the energy released by the extranuclear electrons, then the information ofunderground water could be known through processing by the inversioninterpretation. Considering these advantages mentioned above, MRS method hasbeen widely applied in groundwater exploration, dam leakage, detection of landslidestability and tunnel/mine water bursting, etc.However, the MRS signal(in an order of nv) is so weak that is easy to bedisturbed by random noise, power frequency harmonic noise and pulse noise whichare environmental noise. At this time, the characteristic parameters extracted bycollecting signals are inaccurate which will affect the accuracy of inversioninterpretation. Therefore, how to effectively inhibit interference and improve thesignal to noise ratio in the MRS signal becomes critical.According to this difficulty, based on the summary of denoising methods at homeand abroad for MRS signal and analysis of data characteristic for FID signals andfull-wave MRS signals, a full-wave signal is determined to be the process object inthis paper. Empirical Mode Decomposition (EMD), as a new method fortime-frequency process of signals, has a specialty that is different from these analysistools, such as short-time Fourier transform and wavelet transform, all of which areneed to choose base functions beforehand to decompose signals. Without any priorknowledge of input signals, EMD is able to decompose signals into several Intrinsic Mode Functions (IMF) self-adaptively to extract signal trend effectively. Therefore,this paper proposes the corresponding denoising strategy based on the EMD theory,and verifies reliability and validity of these strategies by simulation and measurementdata.Firstly, it is an important factor to choose the screening cut-off condition thataffects the decomposing result of EMD. From the waveform of reconstructed signalsand improving degree of SNR, SD stop condition and Rilling stop condition based onthe amplitude ratio are analyzed, then the Rilling stop condition is more suitable forMRS signal processed by EMD.Secondly, According to the pulse noise in MRS signal with the characteristic ofwide spectral range,big amplitude and Occasional.it will lead to modal aliasingproblem when use EMD to process pulse noise,that it is difficult to realize the signaland noise.Therefore, this paper proposes the corresponding denoising strategy basedon the Nonlinear Energy Operator (NEO).it not only can deonsing big pulse,but alsocan denosing small pulse.Random noise having zero auto-correlation function atmaximum amplitude, and increases latency close to zero, while the power-frequencyharmonics of the auto-correlation function still has the same periodicity.Based onauto-correlation characteristics of random noise and power frequency harmonics,thispaper presents a denoising algorithm based on EMD and auto-correlationcombining,it Reconstructed signal taking Auto-correlation characteristics of eachsub-band of the IMF after EMD decomposition as a judge to determine thesignal-to-noise criteria.In order to obtain more effective MRS trend term extraction,wavelet thresholding strategy has also been adopted.Finally,a denosing experiment for measured signal was conducted. The resultindicates that this method proposed above can remove noise from full-wave MRSsignal. What is more, signal trend can be picked up efficiently and averagesignal-noise ratio rises by15dB.Compared with conventional filters and classicalwavelet, it can be proved that this method can be useful and effective.