Wi-Fi fingerprint using radio map model based on MDLP and euclidean distance based on the Chi squared test

Abstract

The Wi-Fi fingerprint, which can be used on existing wireless networks, is one of the main indoor positioning techniques that utilizes the received signal strength (RSS). In smartphones, the positioning performance of the fingerprint has been significantly improved through fusion algorithms along with terrestrial magnetism and acceleration sensors. However, the positioning accuracy and speed of the fingerprint is based on radio maps. Although these maps are separate databases obtained without using these sensors, they are important reference elements for initial position estimation and sensor error compensation. In order to minimize the DB of fingerprint and to improve the speed of system construction according to the area of positioning is expanded, this paper proposes a Wi-Fi fingerprint using a radio map construction model based on the minimum description length principle, which can automatically optimize radio maps and the Euclidean distance algorithm based on the Chi squared test. Unlike the existing RSS-classification-based radio map construction method, the proposed access point (AP) classification-based radio construction model not only automatically distinguishes the continuity of all the RSSs acquired from the APs but also optimizes the radio map by eliminating unnecessary APs, based on the information gain. In the positioning phase, based on the proposed radio map, the accuracy of the signals is distinguished using a Chi squared test for the AP RSSs measured in real-time. Therefore, the Euclidean distance, based on the Chi squared test, improves the positioning performance by determining the position accuracy using weighted values of the RSSs, with high reliability.