A Real-Time Sign Language to Text Conversion System for Enhanced Communication Accessibility

Abstract

The research addresses the problem of converting American Sign Language (ASL) finger spelling into text in real-time, enhancing communication for the deaf and hard of hearing. A convolutional neural network (CNN) is utilized to recognize hand gestures from camera images, focusing on the position and orientation of the hand to create accurate training and testing data. The methodology involves filtering hand images, followed by classification to predict the corresponding sign language characters. The calibrated images are then used to train the CNN model. Key findings demonstrate that the proposed system effectively recognizes ASL finger spelling with high accuracy, offering a valuable tool for improving accessibility in communication. These findings suggest significant potential for further applications in real-time sign language interpretation.