Abstract: Industrial frequency noise comes from the electromagnetic noise produced in social activities, and it causes apparent resistivity curves to become pathological or divergent. To improve the accuracy of data processing and interpretation, this study used the Hilbert-Huang transform (HHT) to remove the common power frequency noise in EH-4 data. According to the time series processing and analysis results of measured data, this method can self-adaptively decompose signals according to the time-scale characteristics of the data and successfully remove the industrial frequency noise in the data, thus providing an effective way to remove the noise in magnetotelluric signals. In addition, this study also analyzed the serious modal aliasing and "end effect" occurring in the process of the empirical mode decomposition and decomposed simulation signals and the time series of measured data using the ensemble empirical mode decomposition (EEMD), effectively solving problems such as modal aliasing.