Community detection based reference points clustering for indoor localization in WLAN

For the indoor positioning issue in WLAN based on fingerprinting, reference points clustering methods such as K-means and affinity propagation are frequently used to reduce the region of search. However, the number of clusters needs to be predefined directly or indirectly, meanwhile an unsuitable clustering pattern would lead to poor estimation accuracy, which reduces the practicability of these methods. Based on the theory of the complex network, this paper presents a community detection based reference points clustering for indoor localization in WLAN. A novel clustering target function is proposed and a modified Clauset-Newman-Moore (CNM) algorithm is presented to solve this function. Experimental results demonstrate that, compared to other clustering based localization methods, the proposed method can obtain more accurate estimation.