Abstract: According to the characteristics of quartz micro-electro-mechanical systems (QMEMS) gyro, adaptive filtering and UD decomposition theories, an improved adaptive method based on Kalman filtering method is proposed. This method removes the error data during sampling and enhances the real-time performance and stability of inertial navigation system. The simulation results of QMEMS gyro output show that the improved adaptive filtering is much better than Kalman filtering on de-noise effect. The improved adaptive filter provides more accurate data for inertial navigation system, and the precision of inertial navigation is improved.