Attention-Based Approach for Arabic Sign Language Recognition, Supporting Differently Abled Persons

Sign language serves as a critical communication channel for non-speaking individuals. In this paper, we use a vision transformer (ViT) to classify static images representing the alphabets of Arabic Sign Language. We employ two Kaggle Arabic alphabet datasets, one with over 15,000 static images, und...

Full description

Saved in:
Bibliographic Details
Published inJournal of Disability Research Vol. 4; no. 4
Main Authors Almufareh, Maram Fahaad, Tehsin, Samabia, Humayun, Mamoona, Kausar, Sumaira, Farooq, Asad
Format Journal Article
LanguageEnglish
Published 04.08.2025
Online AccessGet full text

Cover

More Information
Summary:Sign language serves as a critical communication channel for non-speaking individuals. In this paper, we use a vision transformer (ViT) to classify static images representing the alphabets of Arabic Sign Language. We employ two Kaggle Arabic alphabet datasets, one with over 15,000 static images, under varying background conditions, consisting of 31 classes. Our approach emphasizes the inclusivity aspect by aiming to minimize background-related biases and enhance accessibility for persons with a communication disability. Experiments reveal that ViT outperforms traditional convolutional neural networks by achieving a peak accuracy. Cross-data validation results show the reliability, generalizability, and robustness of findings by testing results across multiple, independent datasets.
ISSN:1658-9912
2676-2633
DOI:10.57197/JDR-2025-0586