The primary objective of this project is to convert the Arabic Sign Language to a spoken language, which is the first step on the way to implement a two-way system that also converts spoken language to sign language. This would drastically improve communication between vocal and non-vocal people. Specific objectives include:
a. To provide research work on the automated recognition of Arabic sign language, which- as far as we know-has no research done in its automated recognition. The recognized signs will be converted to spoken words by the computer.
b. To contribute to research directed towards facilitating the communication with physically challenged people.
c. To investigate an unexplored application (sign language recognition) of the support vector machine algorithm that has shown highly promising results in other applications.
d. To attempt to minimize cost and complexity by using off-the-shelf components.
e. To assess the completed system in terms of accuracy, robustness, cost, and expandability.
f. To promote applied research relevant to the Kingdom of Saudi Arabia.
This proposal has been prepared for the 7th annual research program of Prince Salman Center for Disability Research. It is related to the invited research topic: “Use of Computers and Assistive Technologies to Improve Quality of Life for Persons with Disabilities”.
Automated recognition of the sign language greatly facilitates communication with deaf and non-vocal people. It is also important for the development of human-machine interface. This project proposes a system for automatically recognizing the Arabic Sign Language using an instrumented glove as an interface device and Support Vector Machines as a recognition algorithm. After the sign is recognized, a computer will produce a text or a spoken word that corresponds to the sign. This study represents the first attempt to automate the recognition of the Arabic Sign Language using computers. A wide spread of such a system may lead to the unification of the Arabic sign language over the entire Arab World. Support Vector Machine represents the latest and most powerful pattern recognition algorithm in the field that has never been used in the sign language recognition before. It is certain that utilizing advanced gloves to convert the signs to electrical signal and using the most advance method for the recognition would create the first step towards the complete translation of the Arabic sign language to a spoken language.
The project will be conducted with the collaboration of a local institute for deaf students (Al-Amal Institute for Boys in Al-Qateef). Several students will participate in providing the required signs. The investigators have a broad experience in the applications of support vector machines and other artificial neural networks algorithms. The principle investigator is currently supervising a Master thesis conducted on the same topic with encouraging preliminary results even though it utilizes a cheap instrumented glove.

Dr. Talal O. Halawani, PhD; Dr. Samir A. Al Baiyat, PhD; Mr. Jaafar M. Akakah, PhD.

A real time Arabic sign language recognition system that translates singlehanded signs from the Arabic Sign Language Dictionary into speech and text was built. Initially, the system was developed using the Power Glove that was designed for Nintendo games and a support Vector Machine. Due to the limited capability of the Power Glove, the system was able to only recognize about 120 words with accuracy of 70%. Then a system with the CyberGlove and a tracker (Flock of Bird) was developed for 344 single handed signs from the Arabic Sign Language Dictionary using a pattern matching method with an accuracy of approximately 95.7% The Support Vector Machine was used for the recognition of 344 single handed signs from the Arabic Sign Language Dictionary with an accuracy of 98.33%.

King Fahad University for Petroleum & Minerals

N/A.

1. 'Automatic Translation of Arabic Text to Arabic Sign Language', Mohandes, Mohammed, AIML Journal, Volume (6), Issue (4), December, 2006. Thesis by Al- Buraiky,  Salah M.S (2004), 'Arabic Sign Language Recongnition using an Instrumented Glove' UMI Microform 1424619

N/A.

N/A.