Wi-Ear: A Contact-free Vibration Sensing and Identification System Based on COTS WiFi

Mechanical vibration sensing is one of the most critical issues for advanced industrial IoT applications, such as troubleshooting and working condition analysis. Different from status quo solutions, the approaches based on wireless sensing have the advantages of non-invasive and easy to deploy. However, existing wireless works such as RFID and radar are limited by deployment distance or NLoS working scenarios. In this work, we propose a wireless vibration sensing system using COTS WiFi named Wi-Ear, which could monitor multiple devices in LoS or NLoS scenarios. Moreover, Wi-Ear can distinguish the vibration frequency belonging to which device without utilizing any sensor. Especially, in a weak signal scenario where the vibration signal is buried by noise in the raw CSI, Wi-Ear can also detect the vibration frequency effectively. Finally, we implement Wi-Ear with COTS WiFi devices and evaluate it with commercially available motors. For vibration sensing, Wi-Ear can sense weak vibration with an error of 0.1Hz. For device identification, Wi-Ear can identify 10 vibrating devices of 3 different types with an accuracy of 92%. Comprehensive experiments in various scenarios are conducted to show great robustness and stability.