Crossing Linguistic Boundaries: A Unified Real-Time Approach to Multilingual Sign Language Recognition in Diverse Contexts

Abstract

Living a regular life can be challenging for deaf individuals due to their hearing loss. In Bangladesh, around 2.6 million people experience hearing loss. For these individuals, sign language is vital for communication. However, the accessibility, affordability, and availability of traditional sign lang. techniques are frequently restricted. This study introduces a new approach to sign language recognition using the MediaPipe, which helps build machine learning and computer vision systems. Our proposed model effectively recognises letters, sentences, and numbers in both Bengali and English from webcam-captured sign lang motions using ML techniques. Trained on over 135,000 images of hand gestures, the model achieved an accuracy of over 98%. It is appropriate for usage in real-world scenarios since it can adjust to different lighting situations, backgrounds, and hand positions. This technology offers a real-time, affordable, and easily available alternative for sign language interpretation, which has the potential to greatly enhance communication in our nation between the hearing and deaf communities.