American Sign Language Recognition for Underrepresented Populations

Abstract

Interactive educational platforms for learning standardized material, such as new languages or academic topics, have become increasingly popular. However, American Sign Language (ASL) educational tools remain limited, despite the need for accessible and effective ASL learning resources. Artificial intelligence (AI) advancements in interactive educational applications have greatly improved their functionality and versatility. AI is a highly viable and appropriate approach to creating a tool for ASL learning. In translating between text-based languages, there is a simple and consistent mapping between corresponding words and phrases. ASL requires analysis of spatial and temporal features, making AI integration uniquely challenging. This project explores the limitations of ASL education, particularly in the context of interpreter supports and technology. Our project explores various AI models that can effectively promote ASL learning, and provides experimental results for the implementation of various 2D Convolutional Neural Networks (CNNs). Our research prioritizes ethical considerations by carefully selecting datasets to minimize bias, ensuring that AI-driven ASL tools promote inclusivity and accuracy in sign language learning