Towards Activity Recognition Using Wi-Fi CSI from Backscatter Tags

Recently, activity recognition using Wi-Fi CSI has received significant attention due to its low deployment cost. However, its performance depends on the number of Wi-Fi devices, which may be limited in practical scenarios. To address this challenge, we propose a backscatter-based Wi-Fi CSI extraction method using a low-cost backscatter tag. We evaluate the feasibility of using the backscattered CSI for activity sensing. Our experimental evaluation shows that the backscattered CSI can be successfully extracted in various tag locations. Additionally, our experiment with 3 activities shows that there are significant differences in their CSI data, and that they can be classified using machine learning with 94.3% accuracy, which suggests that activity recognition using backscattered CSI data is feasible.