Context-Free Fine-Grained Motion Sensing Using WiFi

Abstract

WiFi-based motion sensing has received a lot of research attention in recent years. Taking advantage of Channel State Information(CSI) collected from physical layer, previous techniques are able to extract useful information from CSI values to infer human movements. However, these works concentrate either on coarse-grained motion sensing or on fine-grained but context-related motion sensing. In this paper, we propose WiTalk, a new fine-grained human motion sensing technique with the distinct context-free character. To profile human motion using CSI, WiTalk generates CSI spectrograms using signal processing techniques and extracts features by calculating the contours of the CSI spectrograms. We verify the proposed technique in the application scenario of lip reading, where the fine-grained motion is the mouth movements. We implement WiTalk on a commercial laptop. Experiment results show that WiTalk can achieve over 92.3% recognition accuracy to discern a set of 12 syllables and 74.3% accuracy to discern a set of short sentences up to six words.