Abstract: In magnetotelluric sounding, the observed electromagnetic field is often distorted because of static shift, increasing the difficulty of data interpretation. In this paper, Daubechies wavelet decomposition method is proposed to recognize measuring point with static shift based on the theory of wavelet analysis. Horst model with inhomogeneous bodies is established and static shift is removed using median filtering method, phase method and wavelet analysis method respectively. The results show that the phase correction result is not stable, relative error has great fluctuation with the frequency; relative error is larger in high frequency than low one for median filtering method, multi scale wavelet analysis correction method is stable, and relative error is not more than 5%, which can effectively suppress the static shift. Finally, static shift is recognized and corrected using the wavelet analysis method for magnetotelluric (MT) data in an area and better results are achieved.