Sign Language Interpretation using Ensembled Deep Learning Models

Abstract

Listen

Translate article

Communication is an integral part of our day-to-day lives. People experiencing difficulty in speaking or hearing often feel neglected in our society. While Automatic Speech Recognition Systems have now progressed to the purpose of being commercially viable, Signed Language Recognition Systems are still in the early stages. Currently, all such interpretations are administered by humans. Here, we present an approach using ensembled architecture for the classification of Sign Language characters. The novel ensemble of InceptionV3 and ResNet101 achieved an accuracy of 97.24% on the ASL dataset.