WiMate: Location-independent Material Identification Based on Commercial WiFi Devices

Abstract

Material identification is playing an increasingly important role in our daily lives such as public security checks. X-ray-based technologies are highly radioactive because they rely on specialized devices to transmit high-frequency signals. Ultrasound-based technologies are cumbersome due to their large size. RF-based approaches necessitate the use of RFID which is usually expensive to be used in home and office environments. To this end, WiFi-based material identification approach has emerged recently as a low-cost yet effective alternative. In this paper, we propose WiMate, a noncontact material identification system leveraging only off-the-shelf WiFi devices. The key enabler of WiMate is a novel theoretical model we build to characterize how the electromagnetic wave decays when penetrating different materials. Our model identifies a unique feature for each material that only depends on the material itself. Consequently, we can leverage this feature coupling with the machine learning techniques for robust and accurate material identification. We prototype WiMate using low-cost commodity WiFi devices and evaluate its performance in real-world. The empirical study shows that WiMate can identify six different materials, i.e., board, paperboard, nickel, wood chip, iron and titanium, with an average accuracy of 96.20%.