Abstract
Through history, humans have used many ways of communication such as gesturing, sounds, drawing, writing, and speaking. However, deaf and speaking impaired people cannot use speaking to communicate with others, which may give them a sense of isolation within their societies. For those individuals, sign language is their principal way to communicate. However, most people (who can hear) do not know the sign language. In this paper, we aim to automatically recognize Arabic Sign Language (ArSL) alphabets using an image-based methodology. More specifically, various visual descriptors are investigated to build an accurate ArSL alphabet recognizer. The extracted visual descriptors are conveyed to One-Versus-All Support Vector Machine (SVM). The analysis of the results shows that Histograms of Oriented Gradients (HOG) descriptor outperforms the other considered descriptors. Thus, the ArSL gesture models that are learned by One-Versus-All SVM using HOG descriptors are deployed in the proposed system.