Abstract: Due to the existence ofnon-line of sight (NLOS) signals, the positioning accuracy of the traditional ultra-wideband indoor positioning method based on Kalman filtering dropped significantly. In response to this situation, a UWB positioning algorithm based on adaptive NLOS signal suppression and Kalman filter (KF) is proposed. The algorithm modeled and analyzed the UWB received signal, and estimated the covariance matrix of the NLOS signal. The covariance matrix was used to "whiten" the received signal. The KF was used to perform indoor positioning under the background of Gaussian white noise. At the same time, the neural network was used to correct the error online to improve the filtering performance. Experimental results show that this method can obtain sub-meter positioning accuracy in NLOS environment, and has strong robustness.