On the Feasibility of Wi-Fi Based Material Sensing

Abstract

Wireless sensing has demonstrated the potential of using Wi-Fi signals to track people and objects, even behind walls. Yet, prior work in this space aims to merely detect the presence of objects around corners, rather than their type. In this paper, we explore the feasibility of the following research question: "Can commodity Wi-Fi radios detect both the location and type of moving objects around them?". We present IntuWition, a complementary sensing system that can sense the location and type of material of objects in the environment, including those out of line-of-sight. It achieves this by sensing wireless signals reflected off surrounding objects using commodity Wi-Fi radios, whose signals penetrate walls and occlusions. At the core of IntuWition is the idea that different materials reflect and scatter polarized waves in different ways. We build upon ideas from RADAR Polarimetry to detect the material of objects across spatial locations, despite mobility of the sensing device and the hardware nonidealities of commodity Wi-Fi radios. A detailed feasibility study reveals an average accuracy of 95% in line-of-sight and 92% in non-line-of-sight in classifying five types of materials: copper, aluminum, plywood, birch, and human. Finally, we present a proof-of-concept application of our system on an autonomous UAV that uses its onboard Wi-Fi radios to sense whether an occlusion is a person versus another UAV.