Device-Free Hand Gesture Recognition System Based on Commercial Wi-Fi Devices

Abstract

Recently, the hand gesture recognition has attracted great interest of researchers due to its important role in human-computer interaction (HCI), smart home applications and virtual reality (VR). The conventional systems mainly utilize additional equipment, such like dedicated sensors and cameras, resulting in higher cost and limitation in application scenarios. In this paper, we present a new gesture recognition system by leveraging the channel state information (CSI) which can be extracted from commodity Wi-Fi device. We design a novel interference elimination algorithm to diminish the influence caused by the signals reflected from static objects and the signal that travels from the transmitter to the receiver directly. After interference elimination, the system can capture the signal reflected from the hand and sample this signal. Then, the sample values are used to construct a virtual antenna array to estimate the moving trajectory of hand. At last, we use Support Vector Machine (SVM) to classify the trajectories and complete the gesture recognition. The extensive analytical and experimental results demonstrate our system can achieve an average accuracy of 0.97 for designed 6 single-hand gestures. Moreover, the system is capable of performing two-hand gesture recognition and it can reach an average accuracy of 0.95 for designed 3 two-hand gestures.