Magnetic Resonance Sounding Oscillating Signal Extraction Based on TLSP Method in the Presence of Co-frequency Noise

Abstract

Magnetic resonance sounding (MRS) is widely employed in groundwater detection, where the parameters of MRS signal are directly linked to the acquisition of hydrological information. However, most traditional MRS signal extraction methods cannot be directly applied to oscillating signals, especially in the presence of co-frequency noise (harmonic noise whose frequency coincides with the Larmor frequency). To address this problem, we propose the Total Least Squares Prony (TLSP) method, which enables the identification and removal of co-frequency noise in MRS oscillating signals. The proposed TLSP method introduces a novel screening mechanism based on attenuation characteristics and frequency characteristics, which successfully eliminates the influence of co-frequency noise. The simulation results demonstrate that TLSP can extract the MRS signals under combined harmonic and random noise interference, even with fundamental frequency fluctuations or multiple harmonic sources. More importantly, TLSP effectively handles co-frequency noise and overcomes the limitation of the conventional Late-time method, which fails for MRS signals with long relaxation times. In 40 experiments under the interference of harmonic noise (S/N = -22dB), random noise (S/N = -8dB), the average S/N_rec of the extracted signal was 36 dB, and in 40 experiments under the interference of co-frequency noise (S/N = -28dB), random noise (S/N = -2dB), the average S/N_rec of the extracted signal was 20dB. The proposed method is also successfully applied to a set of field data.