Weighted Local Access Point based on Fine Matching k-Nearest Neighbor Algorithm for Indoor Positioning System

Demand on Location-based Service (LBS) is rapidly increasing for smart city application. As most of these LBS are acquired from indoor locations, accurate indoor positioning system is important. To date, the most widely used WLAN (Wireless Local Area Network) fingerprint-based i ndoor positioning algorithm is based on state-of-the-art k-Nearest Neighbour (kNN) technique due to its simplicity and robustness. This paper proposes a novel AP weighting technique combined with an improved matching technique for kNN. The strategy is two-fold; first, the signal distance calculation within the algorithm is weighted using signal information of local AP in each fingerprint a nd second, the matching process of the classical kNN algorithm is improved to obtain more accurate position estimates. The results show that the mean error of the proposed algorithm performs up to 14% better than state-of-the-art weighted kNN algorithm and at the same time also outperformed other enhanced version of the kNN algorithm. From the real experiment, we obtained the best k value of 8 which gives the mean positioning error of 2.70 m over an indoor area of 440 square meters.