Cluster filtered KNN: A WLAN-based indoor positioning scheme

Abstract

Location Based Service (LBS) is one kind of ubiquitous applications whose functions are based on the locations of clients. The core of LBS is an effective positioning system. As wireless LAN (WLAN) costs less and is easy to access, using WLAN for indoor positioning has been widely studied recently. K nearest neighbors (KNN) is one of the basic deterministic fingerprint based algorithms and widely used for WLAN-based indoor positioning. However, KNN takes all the nearest K neighbors for calculating the estimated result, which could be improved if some selective work could be done to those neighbors beforehand. In this paper we propose a new scheme called "cluster filtered KNN" (CFK). CFK utilizes clustering technique to partition those neighbors into different clusters and chooses one cluster as the delegate. In the end, the final estimate can be calculated only based on the elements of the delegate. With experiments, we found that CFK does outperform KNN.