It is difficult to establish a precise mathematical model in Extended Kalman Filter (EKF) data fusion algorithm for GPS/DR integrated vehicle navigation system, a novel data fusion algorithm is put forward, which based on wavelet packet analysis. Firstly, the algorithm chooses best wavelet packet group for GPS, odometer and gyroscope signals with least cost principle, and does dynamic pre-filtering for the signals by high frequency coefficients de-noising threshold method based on Stein's Unbiased Risk Estimation (SURE). Finally, the algorithm provides precise positioning information for the GPS/DR integrated vehicle navigation system through doing EKF data fusion. The simulation results show that the algorithm is better than EKF data fusion algorithm, and improve data fusion effect for the GPS/DR integrated vehicle navigation system.