Radio signal based device-free velocity recognition

Abstract

Existing work on RF-based movement recognition focus on analyzing the received signal strength (RSS) in the physical layer. Most of the approaches require the Line-of-Sight signal which limits the application in the Non-Line-of-Sight (NLOS) environment. More importantly, how to distinguish the velocity of human is still an open problem. In this paper, we present an approach for device-free velocity recognition leveraging the radio-frequency (RF) signal in the NLOS environment which is a great challenge for activity recognition. We extract features from the information of received packets to classify different velocities. The classification of human speed can achieve high accuracy in the experiment.