Target Identification, Wi-Fi Imaging, Wireless Sensing, Deep Learning

Identification of Conductive Tags Based on Radio Reflection Patterns Using LSTM

Identifying humans and things (i.e., target identification) is essential for various applications such as inventory management, working time management, and security. RFID is often used for this purpose, but its deployment cost of RFID readers is high. On the other hand, sensing using Wi-Fi (i.e., wireless sensing) has been attracting attention because of its low deployment cost, but target identification is one of the key challenges in wireless sensing because wireless sensing does not assume attaching any device to targets. 
For low-cost target identification, our goal is to design a method for identifying people and things using Wi-Fi without training effort nor additional devices. For this goal, we focus on a Wi-Fi imaging technique that captures a spatial distribution of signal strength received by an antenna array like a camera. We aim to realize target identification by generating a pattern unique to reflected waves using conductive materials such as copper tape and conductive threads that significantly affect on reflected radio signals. Then, we identify the target by capturing the unique reflection pattern by using Wi-Fi imaging. Since the imaging result is greatly affected by multi-path fading, it changes depending on the relative positions between the tag, the transmitter, and the array. Therefore, we assume an environment where a target moves in a pre-defined direction (e.g., a door and a passage) and identify tags by LSTM (Long Short Term Memory) using the time series of the images without their positions. 

Back to Research Themes