Augmentative communication based on realtime vocal cord vibration detection

Abstract

A binary switch based on the detection of periodic vocal cord vibrations is proposed for individuals with multiple and severe disabilities. The system offers three major advantages over existing speech-based access technologies, namely, insensitivity to environment noise, increased robustness against user-generated artifacts such as coughs, and reduced exertion during prolonged usage periods. The proposed system makes use of a dual-axis accelerometer placed noninvasively in proximity of the vocal cords by means of a neckband. Periodic vocal cord vibrations are detected using the normalized cross-correlation function computed from anterior-posterior and superior-inferior accelerometry signals. Experiments with a participant with hypotonic cerebral palsy show the proposed system outperforming a popular commercial sound-based system in terms of sensitivity, task time, and user-perceived exertion.