Intelligent Contactless Gesture Recognition Using WLAN Physical Layer Information

Contactless gesture recognition is an emerging interactive technique in ubiquitous and mobile computing. It combines the linguistics with the wireless signals to analyze, judge, and integrate human gestures by the usage of intelligent algorithms. The existing contactless gesture recognition studies can achieve gesture recognition with the machine learning technologies. But in practice, some objective factors, such as the user's position, the non-line of sight condition, can seriously affect the performance of these gesture recognition systems. In this paper, we propose an intelligent and robust contactless gesture recognition using physical layer information. Instead of the usage of machine learning, we learn the gesture characteristics based on the Fresnel zone model of wireless signals. First, we denoise the collected channel state information (CSI) in a sliding window. Then, we extract the eigenvalues of channel phase information based on Fresnel zone model to depict four basic gestures. The features of gestures are independent of the user's position and the signal amplitude. Finally, common-gesture recognition is achieved based on the decision tree classification. Moreover, we develop a hidden Markov model to achieve the complex-gesture recognition. The extensive experimental results show that our proposed method is position-independent and robust. The accuracy of basic-gesture recognition is as high as 91% on average. And, the accuracy of the complex-gesture recognition is also above 85% on average.