Abstract: In view of the influence of NLOS（Non-Line of Sight） in wireless localization，a multi-scale algorithm for NLOS mitigation is proposed based on the analysis of NLOS error.This algorithm combines multi-scale estimated of the signal with the Kalman filtering，using the low-pass filtering characteristic of wavelet transformation to mitigate the NLOS error in the measurements of TOA/TDOA（Time of Arrival/Time Diff of Arrival），and the method of the Haar wavelet is given.The simulation testing result indicates that this algorithm is very good at mitigating the NLOS error and has a high location accuracy under the different NLOS error model and the different channel environment.