CSI-based passenger counting on public transport vehicles with multiple transceivers

Abstract

Wi-Fi sensing has enabled many applications due to the increasing number of commercial Wi-Fi devices and channel state information (CSI) extraction tools. In this paper, we study the application of CSI-based stationary crowd counting using multiple pairs of transceivers. Specifically, we provide an exact count of the number of passengers on the upper deck of a double-decker bus. Most of the previous solutions count the number of immobile people with a pair of transceivers, which leads to limited sensing scales with the maximum countable number achieved by the state-of-the-art solutions being 15. Indeed, few of these solutions consider the impact of the placement of transceivers on the sensing performance. The major innovation of our work is to identify the optimal topology for multiple receivers based on the Fresnel Zone model to improve the quality of data collection and reduce the overall training effort. We consider the impact of the First Fresnel Zone (FFZ) and the distance between transmitters and receivers when deploying multiple pairs of transceivers. This technique can also be applied to other applications, such as localization, tracking of multiple people, multi-person respiration rate monitoring, etc. The proposed topology was compared with a baseline of a pair of transceivers. Our results show that by properly placing transceivers, the accuracy of counting the exact number of passengers can be improved by more than 11.49%, and the sensing scalability can be extended from 11 to 20 passengers, with an average accuracy of 90.83% with conventional machine learning methods only.