A proposed hybrid sensor architecture for Arabic sign language recognition

Abstract

Sign language recognition is a promising application that breaks the barrier between deaf and normal people. However most researchers in different sign language recognition systems employ a single type of sensors to capture signs. In this paper, hybrid heterogeneous types of sensors are integrated to capture all sign features. Leap motion which is recently available, is customized to capture hands and fingers movements. Two digital cameras are used to capture facial expressions and body movement. The system accomplished 95% recognition accuracy for a dataset consists of 20 dynamic signs due to the additional modules for facial expressions recognition and body movement recognition.