Pedestrian Identification System Based on RFID Signaling and Deep Learning

Abstract

In this article, we propose a new pedestrian identification system based on radio frequency identification (RFID) technology and deep learning algorithms. Our system is designed with a strong focus on privacy protection, in addition to achieving accuracy and reliability. We introduce a 2-D RFID tag array to realize spatial diversity as pedestrians move across the RFID sensor system. The backscattered signal strength indicator (RSSI) and phase angle, affected by the gait and body shape of the pedestrian, carry important personal biometric features for identification. To fully leverage the potential of RFID technology in identification, we propose an innovative neural network model, c3dTAnet, based on attention mechanism and bidirectional long short-term memory (BiLSTM). In the experiment, we have a total of 3060 samples from 50 volunteers and backgrounds to evaluate the performance of the proposed system. The results are promising, with the system achieving 99.1% accuracy in fivefold cross-validation. This demonstrates the significant advantages of our system in both accuracy and training speed over existing pedestrian identification solutions.