Abstract: As a new wireless localization technology, UWB has attracted much attention in the field of indoor localization.In order to improve localization accuracy in ultra-wide band, this paper combines the advantages of Kalman filtering and LSTM networks and proposes a long short-term memory neural network (KF-LSTM) algorithm that incorporates Kalman filtering. Firstly, the UWB timing data is processed by Kalman filtering to weaken the Gaussian white noise in the data, and then the data is put into the LSTM network for training, which takes advantage of the LSTM network's processing of timing features to deal with the non-Gaussian noise and then obtains a more accurate label location.The final measured data show that the average localization accuracy of the KF-LSTM algorithm is improved by 70.21%,37.28% and 38.23% compared to the BP, KF-BP and LSTM network algorithms respectively, and the KF-LSTM algorithm performs more stably.

Key words: ultra wide band, long short-term memory, Kalman filtering, indoor positioning, deep learning