Reducing calibration effort for indoor WLAN localization using hybrid fingerprint database

Abstract

Due to the implementation ease and cost-efficiency, the indoor Wireless Local Area Network (WLAN) fingerprint based localization approach is preferred compared with the conventional trilateration localization approaches. In this paper, we propose a new semi-supervised learning algorithm based on manifold alignment with cubic spline interpolation to reduce the offline calibration effort for indoor WLAN localization using hybrid fingerprint database. The proposed approach significantly reduces the number of labeled training samples collected at each survey location by constructing the hybrid database via interpolation and semi-supervised manifold learning. We carry out extensive experiments in a ground-truth indoor environment to examine the localization accuracy of the proposed approach. The experimental results demonstrate that our approach can effectively reduce the calibration effort, as well as achieve high localization accuracy.