In order to effectively compensate for MEMS (micro electro mechanical system) gyroscope random drift, and improve the carrier attitude estimation’s precision, a wavelet based Kalman filter is proposed.First，the wavelet analysis method is used to do multiscale decomposition with the data of gyroscope to the data signal smooth, then the data analysis of time series modeling to establish the gyro error model is carried on, the low frequency signal of wavelet decomposition is used as the system input, using linear least squares fitting to improve Kalman filtering algorithm , reduce the zero random error of gyroscope. Experiments analyzed the static stationary signals and nonstationary dynamic signal, and made comparative analysis with the treatment effect. The simulation results show that the method can effectively reduce the SNR, improve the accuracy of gyro.