Bridging the Gap: Review of Real-Time Automatic Sign Language Subtitling Using 3D Avatar Technology

Abstract

A large global population experiences some degree of hearing loss, limiting access to multimedia content that depends on spoken language. In Sri Lanka, Deaf and Hard of Hearing individuals communicate primarily in Sri Lankan Sign Language, yet existing subtitle systems fail to convey essential sign language features such as visual grammar, facial expressions, and visual prosody. Although recent advances in automated sign language generation have enabled real-time 3D avatar-based signing for several international sign languages, there is currently no integrated system that converts multimedia subtitles into Sri Lankan Sign Language in real time. This review article examines existing local and international research related to Sri Lankan Sign Language datasets, subtitle processing and synchronization, text-to-gloss translation, avatar-based sign synthesis, motion representation, and user-centered evaluation. The transfer of technology, identification of research gaps, and the existence of systems level requirements that can be met, has been determined through a systematic review methodology. Initial research on gesture modelling and procedural animation has been done by Sri Lanka. However, Sri Lankan researchers have developed very few linguistic resources to support the development of avatar frameworks that allow for a more expressive representation of individuals using avatars. In contrast, studies from other countries demonstrate mature techniques for gloss translation, real-time subtitle alignment, and avatar-based sign language animation that could be adapted for the Sri Lankan context. The review concludes that developing an automated pipeline to translate subtitles into Sri Lankan Sign Language using avatars is both feasible and essential. Such a system would significantly enhance multimedia accessibility and provide a strong foundation for future inclusive technology research in Sri Lanka.