A novel clustering and KWNN-based strategy for Wi-Fi fingerprint indoor localization

With the increasing usage of Wi-Fi infrastructure, methods of indoor localization by Wi-Fi are receiving more and more research efforts in the past. Reducing computational complexity and improving the rate of matching effectively can improve accuracy and real-time of localization. In this paper, we propose a novel clustering approach-AP similarity clustering and K-Weighted Nearest Node (KWNN) method for Wi-Fi indoor localization system. In offline stage, fingerprint in database is trained and divided into different clusters based on different APs´ similarity. In online test stage, firstly, a mobile node finds out the suitable sub-cluster, secondly, we use k-weighted nearest node algorithm to select k nodes among this cluster rather than the whole fingerprint database to reduce the matching space. To validate our strategy, we simulate and compare the proposed approach+KWNN and k-means+KWNN and KWNN-only using the real fingerprint data collected from 900m2 indoor environment. Experiment results reveal that our algorithm will improve localization accuracy by 17.14% and reduce position time-consuming by 50%.