HeadSee: Device-free head gesture recognition with commodity RFID

• Published in: Peer-to-peer networking and applications 2022-05, Vol.15 (3), p.1357-1369
• Main Authors: Chen, Kai, Wang, Fuwei, Li, Miaoyu, Liu, Baoying, Chen, Hao, Chen, Feng
• Format: Article
• Language: English
• Subjects: Body parts; Cameras; Commodities; Communications Engineering; Computer Communication Networks; Engineering; Gesture recognition; Head movement; Human body; Information Systems and Communication Service; Internet of Things; Methods; Networks; Peer to peer computing; Privacy; Radio frequency identification; Sensors; Signal strength; Signal,Image and Speech Processing; Smart buildings
• ISSN: 1936-6442; 1936-6450
• DOI: 10.1007/s12083-021-01126-1

Research shows that the head posture not only contains important interpersonal information but also is an external manifestation of human psychological activities. Head posture plays an important role in automotive safety, smart home, and other intelligent environments. Rf-based posture recognition method provides a non-contact and privacy protection method to detect and monitor human activities. However, how to separate weak activity state information from reflected signals has been a big challenge for this kind of method. This paper proposes HeadSee, a passive human head gesture sensing system built on a cheap commodity RFID device. Without attaching any device to the human body, HeadSee using ICA extracts the weak reflected RF signals from the human body for gesture sensing. And then HeadSee carefully models the head movement by utilizing the signal’s phase/RSS (received signal strength) changes and successfully quantifies the head gesture with continuous sequences of movement states. Extensive experiments show that even with interfering movements from other body parts, HeadSee can still achieve around 91% recognition accuracy of the head gestures.