Wireless Gesture Recognition Based on Multi-cell Cellular Signals

Wireless sensing based on cellular signals has attracted widespread attentions. To achieve two-dimensional (2D) gesture recognition, current researches primarily employ a single-transmitter-multiple-receiver configuration, which however increases the deployment costs and collaboration difficulties among multiple receiving equipments. In fact, cellular signals across various frequency bands and from different operators are filled in the air. If signals from multiple base stations (BSs) can be utilized, there is an opportunity to achieve 2D gesture recognition using only a single receiver, and this will greatly ease the requirement for actual deployment. In this paper, we study a wireless gesture recognition method based on multi-cell cellular signals. First, we discuss how to acquire clean channel state information (CSI) in various multi-cell signal scenarios. Then, without prior knowledge, we propose a scheme to determine whether the BSs of different cells are co-located. For co-located BSs, we combine the CSIs from different bands to enhance the hand reflection path. For non-co-located BSs, the gesture trajectory can be recovered based on Doppler frequency shift information estimated from each transmitter-receiver pair. Finally, we build a prototype system to receive commercial LTE and NR signals, and demonstrate that even a single receiver can achieve excellent gesture recognition performance.