mSense: Towards Mobile Material Sensing with a Single Millimeter-Wave Radio

Target material sensing in ubiquitous contexts plays an important role in various applications. Recently, a few wireless sensing systems have been proposed for material identification. Yet, prior work usually requires to capture the signals penetrating a target (with devices set up on both sides of the target) or to instrument the target (e.g., by attaching an RFID tag), relies on multiple transceivers, and/or involves unexplainable feature engineering. In this paper, we explore the feasibility of material identification by analyzing only the signals reflected off the target, rather than those penetrating it, with a single RF radio. We present mSense, a mobile material sensing system using a single millimeter-wave (mmWave) radio. At the core of mSense is the insight that different materials reflect RF signals in distinct ways. We propose a novel and easy-to-measure material reflection feature that quantitatively characterizes the material's reflectivity. A set of techniques are then devised to achieve accurate and robust material identification despite various factors, including device mobility, hardware defects of commodity mmWave radios, environmental interferences, and etc. Experiments using commercial mmWave networking chipsets demonstrate an average accuracy of 93% in categorizing five common types of materials: aluminum, ceramic, plastic, wood, and water, regardless of their different sizes and thicknesses. The accuracy retains about 90% even in mobile scenarios (i.e., a user holds and moves the radio to perform sensing), which shows the great potential of mSense for mobile applications. A case study on 21 daily objects of various materials, shapes, and textures over different days further validates the performance in differentiating real-life objects.

CCS Concepts: • Human-centered computing → Ubiquitous and mobile computing.