Device-Free Indoor People Counting Using Wi-Fi Channel State Information for Internet of Things

People/crowd counting is a critical technique in many people-centric Internet of Things (IoT) applications, e.g., security monitoring and energy management for smart homes. Device-free people counting systems can in general be categorized as image-based and non-image-based. Non-image-based methods have the advantages of being economical and nonintrusive, as only ambient wireless signals from off-the-shelf wireless devices such as Wi-Fi are used. In this paper, we propose a non-image- based people counting system based on the deep neural network (DNN) model using fine-grained physical-layer wireless signatures such as Wi-Fi channel state information (CSI). Only one Wi-Fi transmitter and one laptop receiver are required, and people are not required to wear or carry any equipment (i.e., device-free). A novel feature space expansion scheme that incorporates the dynamic information of CSI measurements is proposed for the DNN model to enhance its performance. Real testbed experiments showed that the proposed system can achieve as high as 88% average correct classification rate in estimating the exact number of the crowd of size up to nine people in the most general indoor scenario.