In-Air Handwriting by Passive Gesture Tracking Using Commodity WiFi

Recent years have witnessed the great potential of adopting Channel State Information (CSI) for human-computer interaction by gestures. However, most current solutions either depend on specialized hardware or demand priori learning of wireless signal patterns, which face critical downsides in availability, reliability and extensibility. Hence this letter presents AirDraw, a novel learning-free in-air handwriting system by passive gesture tracking using only three commodity WiFi devices. First, we denoise CSI measurements by the ratio between two close-by antennas, and further separate the reflected signal from noise by performing Principal Component Analysis. Besides, we propose a robust signal calibration algorithm for tracking correction by eliminating the static components unrelated to hand motion. The prototype of AirDraw is fully realized and evaluated in real scenario. Extensive experiments yield that AirDraw can track user's hand trace with a median error lower than 2.2 cm.