Fingerprint-based difference-means algorithm for indoor positioning in wi-fi environment

Abstract

In this paper, we propose the fingerprint-based DIFFERENCE-MEANS algorithm (FBDMA) to improve the processing speed and the positioning accuracy in an indoor wireless-fidelity (Wi-Fi) environment. We have applied three design philosophies in the proposed algorithm: the consideration of volume of reference point (RP) in building up signals map, the increase of positioning accuracy by minimum Wi-Fi AP usage, and the increase of positioning accuracy in real-time processing. The architecture of the proposed algorithm consists of two steps: the fingerprint learning step and the indoor positioning step. The main performance metrics in the proposed algorithm are defined as the ratio of successful positioning, the error distances, and the measuring time. The results of the proposed algorithm are summarized as follows: 1) The average ratio of successful positioning in all 24 RPs in the proposed positioning algorithm is measured to 95.8%. This is higher than the 92.7% accuracy achieved by the fingerprint-based algorithm (FBA) and the 80.2% by the fingerprint-based Gaussian-distribution algorithm (FBGA). 2) The average distance of measured errors of the proposed FBDMA is 0.95m; it is less than FBA’s 1.52m and 1.63m of FBGA. 3) The average measuring time of the proposed algorithm is reduced by over 52.3% when compared with FBA and FBGA. In the future, we will upgrade the proposed algorithm that estimates the user trajectory moving continuously in the RPs of a more extended indoor environment, based on this paper, and confirm the predefined performance metrics.