Utilizing wireless signals for gesture recognition and human identification is an emerging type of technology for touchless user interface, which allows the computer to automatically identify the user and interpret his/her gestures as commands. Such techniques extract features to profile the fluctuation of time series wireless signals to infer human gestures/identities. Among which, device-free approach becomes more attractive because it does not need human to carry or wear sensing devices. Existing device-free solutions, though yielding good performance, require heavy crafting on data preprocessing and feature extraction. In this paper, we propose RFnet, a multi-branch 1D-CNN based framework, that explores the possibility of directly utilizing time series RFID signal to recognize static/dynamic gestures as well as the identity of users, which can benefit a large number of applications such as smart homes where security is also a prior concern. We conduct extensive experiments in three different environments. The results demonstrate the superior effectiveness of the proposed RFnet framework.

利用无线信号进行手势识别和人类识别是一种用于非接触式用户界面的新兴技术，它允许计算机自动识别用户并将其手势解释为命令。这样的技术提取特征以剖析时间序列无线信号的波动以推断人类手势/身份。其中，无设备方法变得更加有吸引力，因为它不需要人携带或佩戴传感设备。现有的无设备解决方案尽管产生了良好的性能，但需要对数据预处理和特征提取进行大量设计。在本文中，我们提出了RFnet，这是一个基于多分支1D-CNN的框架，该框架探讨了直接利用时间序列RFID信号识别静态/动态手势以及用户身份的可能性，这可以使诸如智能家居等众多应用程序受益，而安全性也是人们关注的重点。我们在三种不同的环境中进行了广泛的实验。结果证明了所提出的RFnet框架的优越性。