Passive People Counting using Commodity WiFi

Abstract

Indoor people counting is crucial for many applications such as crowd control and smart building. Recent works have shown the potential of using Radio Frequency (RF) signals to estimate the occupancy level. However, most of the existing solutions require training, dense links of many devices, and usually work for only moving human subjects. In this work, we consider people counting in a quasi-static scenario and propose a non-intrusive training-free method using the Channel State Information (CSI) on a single pair of commercial WiFi devices. Different from crowd counting for moving targets that alter the environment significantly, static crowd counting is non-trivial because stationary users only produce minute changes to the wireless signals. First, we transform the quasi-static crowd counting into a continuous multi-person breathing rate estimation problem. Then we propose a novel solution, including an iterative dynamic programming and a trace concatenating algorithm that continuously track the breathing rates of multiple users. By utilizing both spectrum and time diversity of the CSI, our system can correctly extract the breathing traces even if some of them merge together for a short time period. Extensive experiments are conducted in two distinct environments (an on-campus lab and a car). The results show that our system achieves an average accuracy of 86% for both cases. For 97.9% out of all the testing cases, the absolute error of crowd number estimates is within 1.