Aiming at the integrated positioning problem which combines pedestrian dead reckoning(PDR) and global position system(GPS), a wavelet transform-based uscented Kalman filter(UKF) which is used for data fusion is proposed for positioning results from PDR and GPS. Firstly, the PDR/GPS integrated positioning system mathematical model is established. The wavelet transform is adopted to on-line estimation of motion acceleration noise characteristics, and to update the covariance matrix of UKF. The proposed WT-UKF method makes up the traditional UKF’s defects that the filtering effect and accuracy are affected because the signal noise is assumed as Gaussian white noise. Experimental results show that PDR/GPS data fusion with better stability and higher precision are obtained by using the WT-UKF method.