SoundWrite

Abstract

Interacting with explosively growing mobile devices is becoming imperative. This paper presents SoundWrite, a mobile acoustic sensing system that enables text input into commercial off-the-shelf devices without any accessories. SoundWrite leverages the embedded microphone to capture subtle audio signals emitted from writing text on common found surfaces (eg., a wood table). It then extracts distinguishable features from both time and frequency information of received signals to recognize the text. We prototype SoundWrite on Smartphones as an Android application, and perform in-depth evaluation. The evaluation results validate the effectiveness and robustness of SoundWrite, and demonstrate that it could achieve an average recognition accuracy of above 90%.