Reducing Search Area in Indoor Localization Applications

Abstract

In this paper, we propose vertical and horizontal methods to reduce size of search area in indoor localization applications. Although, larger fingerprints means better accuracy, mostly indoor localization applications are running in mobile devices with limited battery, memory, and even processing power. In the proposed approaches, we reduce the size of fingerprints using reducing Access Points (APs) information (vertical reduction) and reducing fingerprint records (horizontal reduction). In vertical reduction, we focus on the importance of APs based on their appearance in fingerprint records. In horizontal reduction, we use regression and decision tree classifiers for primary location estimation. Then, only records in a predefined neighbourhood radius are selected for final localizations. Our studies show that the results of the vertical reduction approaches have a better performance against the results of the horizontal reduction approaches during the indoor localization phase. Also, these findings show that the best way to reduce the size of the fingerprints file is by removing the most common APs from the list.