Hand Movement Velocity Estimation From WiFi Channel State Information

Using WiFi signals for indoor human activity recognition (HAR) has gained popularity in interacting with smart devices. However, prior research has not focused on extracting hand movement velocity during different gestures. Estimating movement velocity is crucial due to individual variations, which impact classifier generalization. Moreover, knowing hand velocity potentially can play a key role in many applications, such as in telemedicine and digital entertainment. This paper proposes a method that extracts Doppler velocity from channel state information (CSI) using MUltiple SIgnal Classification (MUSIC) and simultaneously leverages the information available in multiple access points for estimating hand movement velocity. The findings suggest that while it's feasible to estimate hand velocity with a single access point, employing multiple access points can significantly enhance the accuracy compared to the reference velocities derived from a video camera.