Development of a Sign Language E-Tutor using Convolutional Neural Network

Abstract

Deaf and hearing-impaired people typically use sign language as their primary form of communication. This study designed a Convolutional Neural Network-based Sign Language e-tutor which removes language barriers between people who are deaf and use sign language and people who can hear and speak. Thus giving deaf people a way to communicate with hearing people in real time, with no need to write notes or use a human sign language interpreter. The method used is comprised of four major phases: data collection, data preprocessing, model training and model evaluation. The Model Precision, Accuracy, Recall and f1 score were 0.977, 0.985, 0.99, and 0.99 respectively.