Deep Learning Approach for Arabic Sign Language Alphabet Recognition

Abstract

Introduction: Sign language plays a crucial role in enabling communication for individuals with hearing impairments. Among the various sign languages, Arabic Sign Language (ArSL) is one of the most widely used in the Arab world. It consists of two main forms: word-based ArSL and alphabetic ArSL (ArSLA), where each Arabic letter is represented by a specific hand sign.Objective: This study aims to develop an effective and robust classification model for recognizing Arabic Sign Language Alphabets to enhance communication accessibility for the hearing-impaired community.Method: A Convolutional Neural Network (CNN) architecture was designed and trained on a dataset of Arabic Sign Language Alphabet images. The model’s performance was evaluated using accuracy metrics on both training and testing datasets.Results: The proposed CNN model achieved an accuracy of 99.4% on the training set and 96.57% on the test set, demonstrating its strong generalization ability in recognizing Arabic sign alphabets.Conclusions: The findings confirm the effectiveness of a simple CNN-based approach for Arabic Sign Language Alphabet recognition. This work highlights the potential of deep learning methods to promote accessibility and social inclusion for individuals with hearing disabilities.