Abstract:  In marine controlled-source electromagnetic method (MCSEM) exploration, factors such as seawater motion can cause dipole receiver arms to vibrate, generating dipole vibration noise that reduces the accuracy of subsequent inversion and interpretation. Dipole vibration noise is characterized by its large amplitude, wide frequency band, and short duration. It often manifests as pulse-like distortions, which degrade the waveform, phase, and magnitude versus offset (MVO) curve of MCSEM data. Conventional noise suppression methods are often insufficient for effectively processing this type of noise. This paper introduces the local time-frequency transform (LTFT) based on local smoothness constraint to perform time-frequency analysis on frequency-domain MCSEM data. Leveraging LTFT’s capability for finite discrete frequency point signal decomposition and reconstruction, this method provides low-storage, high-precision time-frequency information representation for MCSEM data with strong time variability. By exploiting the differences in time-frequency domain sparse characteristics between signals and dipole vibration noise, the method separates the noise from the data through time-frequency domain soft-thresholding. This process enables the accurate reconstruction of valid signals. The processing results of synthetic data and field data show that this method can effectively suppress dipole vibration noise, restore the amplitude and phase information of MCSEM signals, make the MVO curve attenuation characteristics more reasonable. Consequently, the proposed approach significantly improves the accuracy of inversion results. 

Key words: marine controlled-source electromagnetic method, dipole vibration, pulse-like noise, soft threshold, local time-frequency transform