Wi-Fi RSS-based Indoor Localization Using Reduced Features Second Order Discriminant Function

There are various methods and technologies for indoor localization. Because of the existence of Wi-Fi access points in many buildings, methods based on Wi-Fi received signal strength (RSS) are the most applicable. In this study, a classifier based on Bayesian decision theory is applied to indoor positioning. The proposed method determines the area where user is located using Wi-Fi RSS classification. The classifier's discriminant function is converted to a second order discriminant function to simplify and reduce computation load. In addition, a feature selection technique based on the area under receiver operating characteristic (ROC) curve is proposed to prevent course of dimensionality problem. The proposed localization system is evaluated by an experimental dataset. The accuracy of 97.9% and 97.45%, using 10-fold cross-validation, are obtained for full features and reduced features, respectively.