Time-domain denoising of time—frequency electromagnetic data

Time—frequency electromagnetic data contain frequency and transient electromagnetic information and can be used to determine the apparent resistivity both in the frequency and time domains. The observation data contains three types of noise: the harmonics interference at 50 Hz, high-frequency random noise, and low-frequency noise. We use frequency-domain bandstop filtering to remove the harmonics interference noise, segmentation and extension median filtering, and fitting of fixed extremes in empirical mode decomposition to remove the high-frequency and low-frequency noise, respectively; furthermore, we base the selection of median filtering window size on the variance and skewness coefficient of the data. We first remove the harmonics interference at 50 Hz, then the high-frequency noise, and finally the low-frequency noise. We test the proposed methodology by using theory and experiments, and we find that the three types of noises are removed, the phase and amplitude information of the signal are maintained, and high-quality waveforms are obtained in the time domain.