Discriminative Learning-based Smartphone Indoor Localization

摘要: Due to the growing area of ubiquitous mobile applications, indoor localization smartphones has become an interesting research topic. Most current systems rely on intensive site survey achieve high accuracy. In this work, we propose efficient system that is able reduce effort while still achieving Our built by fusing a variety signals, such as Wi-Fi received signal strength indicator, magnetic field and floor plan information in enhanced particle filter. To stable performance, first apply discriminative learning models integrate readings room level landmark detection. Further, detection, range-based models, with graph-based discretized state representation. Because our approach requires only learning-based detections, time spent phase significantly reduced compared traditional fingerprinting or landmark-based approaches. We conduct experimental studies evaluate office-like environment. Experiment results show can efforts, method performance average error 1.55 meters.