Abstract

vision basedapproaches are susceptible to the light condition and sensor based methods havedisadvantages in deployment and high cost. The latest researches havedemonstrated that the pervasive wireless signals can be used to identifydifferent gestures. In this paper, we attempt to utilize CSI (channel stateinformation) derived from wireless signals to release the device-free air-writerecognition called WiFi. Compared to the gesture recognition, the increaseddiversity and complexity of characters of the alphabet makes it challenging. ThePCA (Principle Component Analysis) is used for diagnosing effectively and theenergy indicator derived from the FFT (Fast Fourier Transform) is to detectaction continuously. The unique CSI waveform caused by unique writing patternsof 26 letters serves as feature space. Finally, the HMM (Hidden Markov model)is used for character modeling and classification. We conduct experiments inour laboratory and get the average accuracy of the Wri-Fi are 86.75% and 88.74%in two writing areas, respectively.

Keywords: AugmentedReality, Virtual Reality