Wi-PC: A Device-Free People Counting System Based on WiFi Channel State Information

In recent years, wireless sensing based on WiFi has become a popular research topic in recent years. This paper proposes a people counting framework named Wi-PC, which utilizes WiFi information to count people in enclosed space. In this framework, two new features, the bandwidth feature based on the subcarrier domain (BWS) and the Bayesian discovery based on the time-frequency domain (BaTF), are first proposed, respectively. A feature set is then created by fusing BWS, BaTF, and statistical features extracted from the time domain. Finally, the support vector machine (SVM) is used to perform people counting. Experimental results show that the SVM based on the fused feature set achieves an accuracy exceeding 95%, substantially improving the recognition accuracy of the model.