Accurate WiFi Localization by Fusing a Group of Fingerprints via a Global Fusion Profile

The existing indoor localization approaches based on single fingerprints, such as received signal strength (RSS) and channel impulse response, are rather susceptible to the changing environment, multipath, and nonline-of-sight. It is well known that indoor localization can obtain higher positioning accuracy than the single-fingerprint-based methods by fusing multiple information sources (fingerprints/fingerprint functions). However, the existing fusion methods cannot fully exploit the intrinsic complementarity among multiple information sources and thus show lower accuracy. In this paper, we propose an accurate WiFi localization approach by Fusing A Group Of fingerprinTs (WiFi-FAGOT) via a global fusion profile (GFP). WiFi-FAGOT first constructs a WiFi-based GrOup Of Fingerprints (GOOF) in the offline phase, which consists of RSS, signal strength difference, and hyperbolic location fingerprint. Then, instead of direct localization by using the WiFi-based GOOF, we design multiple classifiers by training each fingerprint in the WiFi-based GOOF, namely GOOF classifiers. To fully leverage the intrinsic complementarity among different kinds of fingerprints, we propose a GFP construction algorithm by minimizing the average positioning error over the space of all GOOF classifiers. Finally, in the online phase, we derive a grid-dependent matching algorithm, namely, optimal classifier selection, to intelligently choose a fusion profile in the GFP for more accurate localization. Experimental results demonstrate that WiFi-FAGOT performs better than other systems in real complex indoor environments.