| In this thesis we present a framework for recognizing Arabic Sign Language (ArSL). Sign languages are the primary mode of communication for many hearing impaired people, just as speech is for vocal people. Sadly, the current state of the art in sign language recognition research is still lagging far behind speech recognition. Although some work has been done with American, Chinese, Taiwanese and Greek sign language recognition, there has been very little contribution in ArSL recognition area. This provides the motivation for this thesis, in which we present a vision/camera based system for isolated ArSL recognition for a vocabulary of 300 words (which include both one handed and two handed signs, stationary and moving sings) using Hidden Markov Model (HMM)-based framework for recognition. The idea is then extended to an integrated system which combines both a vision based system and a virtual reality glove based system. The results of the recognition system are reasonably good for both systems. While the accuracies for vision based system ranged from 87% to 96%, the integrated system gave better accuracies ranging from 91% to 99.2%.;Key words. Arabic sign language, hidden Markov Models (HMM). |
