Machine Learning and Deep Learning Approaches for Arabic Sign Language Recognition: A Decade Systematic Literature...

Area of Science:

• Computer Science
• Artificial Intelligence
• Human-Computer Interaction

Background:

• Arabic Sign Language (ArSL) is crucial for communication among Arabic speakers who are deaf or hearing-impaired.
• Effective Arabic Sign Language Recognition (ArSLR) tools are needed to bridge communication gaps, especially for non-signers.
• A systematic review of machine learning (ML) and deep learning (DL) methods for ArSLR is currently lacking.

Purpose of the Study:

• To provide a comprehensive overview of ArSL recognition research.
• To analyze ML/DL methods and techniques used in ArSLR systems.
• To identify challenges and future research directions in ArSLR.

Main Methods:

• A systematic literature review of ArSLR research published between 2014 and 2023.
• Searches conducted across three major databases: Web of Science (WoS), IEEE Xplore, and Scopus.
• Adherence to PRISMA guidelines for study screening, inclusion, and exclusion criteria.

Main Results:

• Analysis of 56 included articles, focusing on datasets and ML/DL techniques.
• Predominance of vision-based approaches in ArSLR research.
• Most studies concentrate on fingerspelling and isolated word recognition, with limited work on continuous sentence recognition.

Conclusions:

• Current ArSLR research is heavily skewed towards isolated sign recognition using visual data.
• Significant challenges remain in developing systems for continuous ArSL recognition.
• Future research should prioritize continuous ArSL recognition and explore diverse methodologies beyond vision-based approaches.