Search

Due to non-line-of-sight propagation and other effects, the accuracy of indoor positioning based on location fingerprint is not high. Aiming at this problem, this paper proposed a novel indoor positioning algorithm, which based on the reliable AP selection and deep belief network ( DBN) . Firstly, the algorithm used the improved K-means clustering algorithm to divide the locating area into several sub-regions in the off-line phase. Then according to the Fisher criterion and the AP absent frequency, it selected the strongly distinguishable and reliable AP node as the training node of the sub-region. And finally it used the DBN model to train the data of each sub-region. In the on-line phase, the improved algorithm determined the cluster according to the received signal strength, and estimated the location of test point by the trained DBN model. The experimental results show that, compared with WKNN, M-WKNN and PSO-ANN algorithm, the proposed algorithm can effectively improve the accuracy and stability of positioning. [ABSTRACT FROM AUTHOR]