Watching Your Phone's Back

Lei Wang,Xiang Zhang,Yuanshuang Jiang,Yong Zhang,Chenren Xu,Ruiyang Gao,Daqing Zhang

DOI: https://doi.org/10.1145/3463522

2021-01-01

Proceedings of the ACM on Interactive Mobile Wearable and Ubiquitous Technologies

Abstract:Gesture recognition on the back surface of mobile phone, not limited to the touch screen, is an enabling Human-Computer Interaction (HCI) mechanism which enriches the user interaction experiences. However, there are two main limitations in the existing Back-of-Device (BoD) gesture recognition systems. They can only handle coarse-grained gesture recognition such as tap detection and cannot avoid the air-borne propagation suffering from the interference in the air. In this paper, we propose StruGesture, a fine-grained gesture recognition system using the back of mobile phones with ultrasonic signals. The key technique is to use the structure-borne sounds (i.e., sound propagation via structure of the device) to recognize sliding gestures on the back of mobile phones. StruGesture can fully extract the structure-borne component from the hybrid Channel Impulse Response (CIR) based on Peak Selection Algorithm. We develop a deep adversarial learning architecture to learn the gesture-specific representation for robust and effective recognition. Extensive experiments are designed to evaluate the robustness over nine deployment scenarios. The results show that StruGesture outperforms the competitive state-of-the-art classifiers by achieving an average recognition accuracy of 99.5% over 10 gestures.