An Improved Location Estimation Method for Wifi Fingerprint- based Indoor Localization

The Accurate indoor localization is a challenging task due to the absence of GPS. Among numerous proposals, Wifi fingerprint-based localization is one of the most promising approach, since most buildings are nowadays equipped with Wifi access points for wireless network coverage. Due to the nature of Wifi access points in which any user can deploy and manage their own, fingerprints from some access points lead to estimation errors. Location estimation algorithms should consider these factors and be able to locate users with low error distance. Finding the nearest neighbor using Euclidean distance in signal space is most widely used method in estimating location. However, this paper shows that Euclidean distance is prone to error when unstable access points are present. Also, Euclidean distance does not differentiate strong signals and weak signals, which can also mislead location estimation. We propose a different way to determine the nearest neighbor, which penalizes signals from unstable access points, and signifies strong signals compared to weak signals. Experiments with real measurements show that the proposed algorithm reduces mean error distance by 57 % and 90-percentile error distance by 64 % compared to the Euclidean distance method