Detecting Pedestrians Behavior in Building Based on Wi-Fi Signals

Human detection and tracking is a key and fundamental problem of smart devices and intelligence applications. Currently, video surveillance and infrared image surveillance are the mainstream solutions of this problem, however suffering from multiple complex difficulties. Meanwhile, the high penetration rate of Wi-Fi equipment and smart phones and mature technology of Wi-Fi positioning make it possible to trace people with Wi-Fi signals. In our approach, a framework has been proposed to recognize people behaviors of "entering" and "egressing" of a building based on Wi-Fi signals. The framework includes three stages: data preprocessing, time slice processing and two-level classification, which contains a One-Class SVM and a Binary-Class SVM. Further analysis of individual and macroscopic behaviors based on the framework output has been made, which can guide the security guard, infrastructure arrangement and energy conservation of a building so as to implement the smart building concept in urban life.