The growing PHY layer capabilities of WiFi has made it possible to reuse WiFi signals for sensing. Sensing via WiFi enables remote sensing without wearable sensors and contactless sensing in privacy-preserving mode, which are beneficial in various applications including security surveillance, intrusion detection, elderly monitoring, and human-computer interaction. For WiFi sensing to excel indoors, multipath propagation acts as a major concern. The multipath effect can invalidate theoretical propagation models, distort received signal signatures, and constrain the performance of wireless sensing even when inferring the presence of humans. To explicitly eliminate any adverse impact of multipath propagation, researchers resort to customized signals and specialized software-defined...[ Read more ]

The growing PHY layer capabilities of WiFi has made it possible to reuse WiFi signals for sensing. Sensing via WiFi enables remote sensing without wearable sensors and contactless sensing in privacy-preserving mode, which are beneficial in various applications including security surveillance, intrusion detection, elderly monitoring, and human-computer interaction. For WiFi sensing to excel indoors, multipath propagation acts as a major concern. The multipath effect can invalidate theoretical propagation models, distort received signal signatures, and constrain the performance of wireless sensing even when inferring the presence of humans. To explicitly eliminate any adverse impact of multipath propagation, researchers resort to customized signals and specialized software-defined radios for radar signal processing. To enable device-free applications on commodity infrastructures, existing approaches exploit a dense deployment of wireless links.

Instead of avoiding multipath, in this thesis, we demonstrate it is possible to harness multipath in WiFi sensing with the PHY layer Channel State Information (CSI). First, we design a primitive to identify the availability of the LOS path under multipath propagation with only commodity WiFi devices to improve the multipath awareness in WiFi sensing. Second, we exploit the rich multipath effect as fingerprints to blur the directional coverage of traditional passive human detection architecture to achieve omnidirectional coverage. Third, we propose a measurable metric as proxy for detection sensitivity and a lightweight subcarrier and path configuration scheme to adapt to different multipath propagation conditions. Finally, we design a unified framework for both static and moving human detection, by capturing the chest movements of static humans. We prototype the above schemes with commodity WiFi infrastructure, and evaluate their performances in typical office environments. Experimental results demonstrate improved detection accuracy, coverage and sensitivity compared with MAC layer RSSI based schemes.