KLDLoc: A WiFi Fingerprint-based Localization Method Based on Kullback-Leibler Divergence

In recent years, with the gradual popularity of wireless local area networks (WLANs) and mobile devices, WiFi fingerprint-based localization technology has won wide attention from researchers by virtue of its simple principle and pervasiveness. This paper discusses one of the key problems in WiFi fingerprint-based localization, namely improving the localization performance with a few samples. To solve this problem, the Kullback-Leibler Divergence(KLD) of two Gaussian distributions is proposed to measure the difference between two fingerprints. From the theoretical analysis, it can be obtained that the proposed one is more robust compared to the mean or variance-based measures. In this paper, we present a Kullback-Leibler Divergence-based solution, termed KLDLoc. Extensive experiments using real-world datasets are conducted and show that in comparison with several existing methods, KLDLoc can improve the localization accuracy by at least 38%.