A Scalable Deep Neural Network Architecture for Multi-Building and Multi-Floor Indoor Localization Based on Wi-Fi Fingerprinting

摘要

One of the key technologies for future large-scale location-aware servicescovering a complex of multi-story buildings --- e.g., a big shopping mall and auniversity campus --- is a scalable indoor localization technique. In thispaper, we report the current status of our investigation on the use of deepneural networks (DNNs) for scalable building/floor classification andfloor-level position estimation based on Wi-Fi fingerprinting. Exploiting thehierarchical nature of the building/floor estimation and floor-levelcoordinates estimation of a location, we propose a new DNN architectureconsisting of a stacked autoencoder for the reduction of feature spacedimension and a feed-forward classifier for multi-label classification ofbuilding/floor/location, on which the multi-building and multi-floor indoorlocalization system based on Wi-Fi fingerprinting is built. Experimentalresults for the performance of building/floor estimation and floor-levelcoordinates estimation of a given location demonstrate the feasibility of theproposed DNN-based indoor localization system, which can provide nearstate-of-the-art performance using a single DNN, for the implementation withlower complexity and energy consumption at mobile devices.