Extracting location information from RF fingerprints

Location Fingerprinting (LF) is a promising localization technique that enables many commercial and emergency location-based services (LBS). Significant efforts have been invested in enhancing LF using advanced machine learning methods. Most of these techniques require a huge amount of geo-tagged training data to achieve significant improvement in accuracy. This increases calibration efforts and cost. In this paper, enhancing the localization accuracy by providing more reliable input data to the LF algorithms is discussed. A method to extract the most information content from the fingerprint measurements is proposed. The localization accuracy is improved without increasing the calibration costs or computational complexity. This solution can be scaled for high volume commercial LBS applications. We prototyped our proposed solution to locate users in an enterprise building scenario. Android mobile users connected to our local cloud localization server are accurately located within the building.