Harvesting indoor positioning accuracy by exploring multiple features from received signal strength vector

Abstract

The development of an indoor location information system using ubiquitous resources available in the environment is a challenging problem in the field of Geo-Location technologies, these days. Therefore, instead of relying on a single resource, the fusion of location information from multiple resources into an indoor positioning system (IPS) becomes important. The IPS in which information from multiple sources such as Wi-Fi, geomagnetism, and motion sensors is fused to harvest the next level of accuracy is commonly known as hybrid IPS. The initial estimate of the position with high accuracy is very critical for the hybrid IPS. Wi-Fi fingerprinting is one of the potential candidates for providing the initial position in such systems, whereas due to the multipath, absorption, and fading characteristics of the indoor environment, the accuracy of the Wi-Fi fingerprinting techniques is limited. Many algorithms and techniques have been proposed to improve the accuracy of Wi-Fi-based IPSs. However, most of the solution requires high computing resources and specialized hardware. This article proposes an empirical approach in which the important features present in the received signal strength vector (RSSV) of the Wi-Fi device are selected to exploit the similarity measure and index order of the Access Points (APs). The experimental results show that these features make it possible to avoid long distances outliers and to improve the positioning accuracy of the Wi-Fi fingerprinting technique without the use of specialized hardware.