Towards Ubiquitous WiFi Sensing through Exploiting Multi-Source Signals

WiFi has been widely used for local area networking of devices and Internet access in the past two decades. Many researches further use WiFi signals for target sensing through analyzing the CSI (Channel State Information) of signals affected by the target movement. These works cannot achieve ubiquitous WiFi sensing as they are implemented based on specially deployed WiFi devices while CSIs are extracted from specific probing packets. ISAC (Integrated Sensing and Communication) is considered to be a future direction to achieve ubiquitous WiFi sensing through extracting CSIs from communication WiFi packets for target sensing. However, current works are all implemented under the single-source signal scenario, resulting in limited sensing range and performance. In this paper, we utilize the existing multi-source signals in the WiFi network to improve the sensing performance. Specifically, when multi-source signals are affected by a target activity, the extracted multiple CSIs can be considered as multiple views for the same activity. We propose a feature fusion process based on DCCA (Discriminant Canonical Correlation Analysis) to fuse the CSIs from multiple views, so as to achieve higher classification accuracy. We implement the system based on USRP platform with two signal sources and conduct hand gesture recognition on the system. Experimental results demonstrate that the recognition accuracy with the single-source signal can be improved by about 10% through feature fusion.