عنوان مقاله [English]
Magnetic Resonance Sounding (MRS), as a surface geophysical method, provides good information about the hydro-geophysical parameters (such as water content and hydraulic conductivity) of aquifers. The main advantage of the MRS method compared to other geophysical methods is that the surface measurement of the MRS signal responds directly to the presence of water below the surface. Despite the high efficiency of this method, the recorded signal is strongly affected by electromagnetic noises, including spike noises and harmonic noises. The first generations of MRS instruments were single-channel instruments. In single-channel instruments, both magnetic resonance excitation and signal recording are done with a single loop, and it is necessary to use various forms of filtering to eliminate noise, in particular powerline harmonics.
Then a new generation of multichannel MRS instruments with multiple is was built, so that the main loop is still used for magnetic resonance excitation and signal recording. In addition, a number of reference loops, physically displaced from the main loop, measure only noise. Parts of the noise recorded by the reference loops can correlate with the noise in the main loop. With proper signal processing, the noise in the reference coils can be filtered to obtain a replica of the noise in the main coil and when this replica is subtracted from the signal recorded in the main loop, hence, the MRS signal remains noiseless. One of the current challenges of MRS signal processing is the existence of harmonic noise with the variable fundamental frequency. If the harmonic signal from a specific source has a fundamental frequency that varies with time, most of the proposed algorithms will not perform well in eliminating harmonic noise. Therefore, a new topic that is followed in this paper is to evaluate the performance of the proposed algorithm in cases where the harmonic signal has a fundamental frequency that varies with time.
In this paper, in order to obtain an accurate estimate of the parameters of the magnetic resonance sounding signal, a method for eliminating spike and then harmonic noise in the time domain is presented. Synthetic signals are contaminated with different electromagnetic noises to investigate the effect of different optimal filter parameters for spike and harmonic event elimination methods. Spike noise has a detrimental effect on the performance of the harmonic noise elimination algorithm. Hence, spike signals must be deleted or adjusted before applying the harmonic noise elimination algorithm. First, a statistical processing algorithm based on the signal-dependent rank-order mean (SD-ROM) filter for eliminating spike noise is presented and after deleting spike noise, a method for eliminating harmonic noise is assumed based on the fixed and variable fundamental frequency with time using remote reference loop. Numerical results of applying the proposed processing algorithms in the time domain show that by applying the mentioned methods, a considerable amount of spike and harmonic signals are removed leading to a good estimate of the parameters of the magnetic resonance sounding signal (i.e., initial amplitude, relaxation time, phase and frequency of the signal).