Bidirectional Language Agnostic Framework for Sign Language Production and Recognition

Abstract

This thesis explores the development of an automated SL production system using cuttingedge advancements in natural language processing (NLP) & computer vision. Motivated by the growing need for inclusive communication solutions, especially for the deaf & hardof-hearing community, the project focuses on generating SL poses from input text. The proposed system leverages a series of interconnected processes, including semantic & grammar correction, translation, word validation & substitution, pose generation, & pose stitching. By ensuring language agnosticism & adaptability across multiple languages, the system aims to bridge communication gaps for deaf individuals in various social, legal, & corporate settings. This research is grounded in an extensive review of the current methodologies in SL recognition & production, highlighting the limitations & potential improvements in existing systems. The final evaluation of the system demonstrates promising results in producing accurate & comprehensible SL videos, contributing to the ongoing efforts to promote accessibility & inclusivity for the deaf community.