Wi-PIGR: Path Independent Gait Recognition With Commodity Wi-Fi

Wi-Fi based gait recognition has many potential applications. However, the information derived from changes with walking path. This makes human identification through really challenging, existing systems require subject along a predetermined path dependence restriction impedes being widely used. In this paper, independent system for single subject, Wi-PIGR, is proposed. identified regardless of Specifically, an extra receiver introduced to get CSI data in orthogonal directions. A series signal processing techniques are proposed eliminate differences among signals by arbitrary paths and generate high quality spectrogram. Furthermore, deep learning approach integrated into feature extraction. The experiment results typical indoor environment demonstrate superior performance average accuracy 77.15 percent, when number subjects 50.