Passive WiFi fingerprinting method

Abstract

WiFi fingerprinting methods are widely used in indoor positioning field, but it requires time and efforts to collect fingerprints. Crowdsourcing techniques have been actively studied to reduce the collection cost, but it still needs user's explicit involvement such as installing and operating an application. In this paper, we propose a network fingerprinting method without the explicit involvement. it collects unlabeled fingerprints including received signal strength(RSS) of probe request message(PRqM) by multiple APs. After collecting the fingerprints, we perform latent semantic analysis(LSA), singular vector decomposition(SVD) and location optimization to construct radio map. The proposed method achieved 2.93m accuracy of radio map and 3.72m accuracy of positioning.