An experimental study of indoor RSS-based RF fingerprinting localization using GSM and Wi-Fi signals

Abstract

Localization of mobile users in indoor environments has many practical applications in daily life. In this paper, we study the performance of the received signal strength (RSS)-based radio frequency (RF) fingerprinting localization method in a shopping mall environment considering both calibration and practical measurement cases. In the calibration case, the test data for the RSS fingerprinting database are built offline by receiving signals from Global System for Mobile Communications (GSM) base stations, which are collected by a dedicated measurement tool, i.e. the Test Mobile System. In order to see the localization performance, the k-nearest neighbors (K-NN) and random decision forest (RDF) algorithms are implemented. The RDF algorithm provides a better localization performance than K-NN in this case. For the practical implementations, the RSS values of both GSM and Wi-Fi signals are collected by ordinary smartphones. Localization is performed using different classification algorithms, i.e. BayesNet, support vector machines, K-NN, RDF, and J48. Moreover, the effects of the received signal type, phone type, and number of reference points on localization performance are investigated.