Proposition of a neural solution to translate sign language into Algerian dialect

Abstract

Sign language is a distinct form of communication essential for various segments of society. It encompasses a diverse range of signs, each characterized by variations in hand shape, motion profile, and the positioning of hands, face, and body parts. Consequently, visual sign language recognition represents a complex area of research within computer vision. In recent years, significant advancements have been made, mainly through using deep learning approaches, as proposed by various researchers. This work focuses explicitly on translating American Sign Language (ASL) into the Algerian dialect, with the overarching goal of bridging the communication gap between the ASL-based deaf community and speakers of the Algerian dialect. The project consists of two primary components. Firstly, a sign language recognition phase, where two models have been developed to detect ASL signs in static images and in real-time accurately. Secondly, a translation phase that employs a word to word translation techniques to convert the recognized signs into the Algerian dialect. key words: Sign Language, Algerian Dialect, Machine Translation ,Computer vision,Deep learning.