Wavelet analysis technique is proposed in the filter processing to remove the noise from random drift signal of fiber optic gyroscope (FOG). Based on the component elements of random drift signal of FOG, mathematical model of random drift signal of FOG is built. According to the shortcomings of hard threshold and soft threshold, threshold function of forced denoising is used in signal filtering. It is shown from simulation and statistic analysis that the wavelet analysis technique is an effective method to eliminate noise from random drift signal of FOG, and that selection of wavelet base and decomposition scale also has an impact on filtering effect of wavelet analysis.