AI-Enabled Sign Language Recognition For Inclusive Education: A Comparative Analysis Of India And Australia Through Computer Vision And Natural Language Processing

Abstract

Over the last decade there has been rapid convergence of artificial intelligence, computer vision and natural language processing which has presented a significant opportunity to transform communication and educational accessibility for deaf learners over the period of time, yet the sign language technologies remain substantially underdeveloped compared with spoken language AI. In reference to India, Census 2011 has recorded 1.26 million persons with hearing disability, while contemporary research estimates that almost 5 to 6 million users of Indian Sign Language, has reflected a substantial population whose linguistic and educational needs cannot be adequately captured through conventional speech-oriented technologies. There has been a constant technological gap which is evident in the limited availability of high quality, representative datasets that is the INCLUDE data set contains approximately 0.27 million frames across 4287 videos covering 263 ISL signs While recent research has continued to identify data set scarcity dialectal variation and infrastructure constraints as barriers to continuous ISL recognition. However, Australia over the period of time has presented a useful comparative setting, with 16,245 people reporting Auslan use at home in the 2021 Census and 2022 National Deaf Census generating 1215 responses, including 846 Deaf, Deafblind, Deaf-disabled and Hard of Hearing Auslan users; Father there has been an establishment of Auslan corpus comprising approximately 300 hours of videos involving 100 narratives and near native singers. Against all this background has evolved for the study making a comparatively understanding how the AI has enabled sign language recognition through computer vision deep learning and NLP, focusing on data set adequacy, recognition, translation, linguistic diversity, educational integration and human AI interaction. Moreover, this shall argue over the technological accuracy alone is however not sufficient and that the inclusive requires representative datasets, culturally responsive modelling, Privacy safeguards and reproducible evaluation. This however will identify different strategies for developing trustworthy, real time and human cantered AI system mechanism which can be capable of strengthening communication and inclusive education for deaf learners worldwide.