ASL2ESL-DL: Arabic Sign Language to English Sign Language translation system using Deep Learning (Prof. Ayed Salman)

Deaf and hearing-impaired individuals use sign language as a form of communication. Sign language involves a series of movements, gestures, facial expressions, and postures that correspond to spoken language letters and words. Various sign languages are used worldwide, with the World Federation of the Deaf reporting in 2018 that more than 70 million deaf individuals use over 300 different sign languages globally. The world currently has different sign languages depending on the geographical location and the standard spoken language used in different countries. This created another layer of difficulty of communications among deaf people with each other’s especially when they are living in different countries. For these reasons, scientists under the World Federation of Deaf (WFD) are currently trying to develop a universal sign language, called 'International Science Sign Language', for faster and easier communication between deaf people and to achieve the principle of equality. Designing a system that can translate between multiple sign languages would also work toward the same goal of easing communication among deaf individuals. This research objective is to enhance communication between deaf people using Arabic Sign Language and those who are using English Sign Language by proposing, designing, and implementing a novel middleware system that uses convolutional neural networks to translate between both sign languages. The proposed system is named Arabic Sign Language to English Sign Language translation system using Deep Learning (ASL2ESL-DL). In this thesis, we designed, implemented, and tested ASL2ESL-DL on the “ArASL2018 dataset, which contains 54,049 images of 32 Arabic alphabet signs gathered from 40 different participants. The proposed model was implemented and showed an accuracy of 97.3% and a total loss of 12%.

Supervisor: Prof. Ayed Salman 
Co-Supervisor: Dr. Mohammed Alenezi 
Convener: Dr. Sa'ed Abed 
Examination Committee: Dr. Sa'ed Abed & Dr. Abdullah Alshaibani