Construction and Management of Fingerprint Database with Estimated Reference Locations for WiFi Indoor Positioning Systems

摘要

In this paper, we propose a positioning system for indoor environments based on WiFi fingerprint methods. In fingerprint methods, the database construction is important for the position estimation performance. We created the fingerprint database by gathering pairs of media access control (MAC) addresses and received signal strength indicator (RSSI) at each estimate reference location. In this proposed method, reference devices are moving at a constant speed with a simple direction that can be estimated the location of each reference point. The multiple data sets are collected in same areas to create the fingerprint database. The location of reference points are slightly different with each other. We use the mean-shift clustering algorithm to get merged reference points from the multiple data sets. We manage to update the database using the user's input data in different time duration. We created four types of fingerprint database and evaluated the user's position estimation. The results show that the user's position can be estimated with an average accuracy error of 1.8 m.