WiCount: A Deep Learning Approach for Crowd Counting Using WiFi Signals

Abstract

The ubiquitous WiFi devices and recent research efforts on wireless sensing have led to intelligent environments which can sense people's locations and activities in a device-free manner. However, current works are mostly designed for single human environment owing to the complexity of multiple human environment and in turn greatly hinder this technology from real implementation. To realize such device-free sensing in a multiple human environment, the first step-stone is to estimate how many targets or in other words crowd counting, which is not only the basis for multiple human environmental sensing but also lead to many potential applications such as crowd control. Previous efforts for crowd counting using WiFi failed to do so, as the robustness of their method is limited. To this end, we propose WiCount - the first solution using a deep learning approach to infer the number of people robustly in the room with WiFi signals. Our scheme is based on the key intuition that now that it is too complex to model the crowd counting using WiFi directly, we can use deep learning approaches to construct a complex function to fit the correlation between the number of people and Channel State Information (CSI) values. The prototype of WiCount is implemented and evaluated on the commercial WiFi device. The experimental results show that our deep learning model is able to estimate the number of crowd up to 5 with the accuracy of 82.3% in a rather effective and robust manner.