Indian Sign Language Interpretation using Convolutional Neural Networks

Sign language is the most common means of communication among the speech- and hearing-impaired. Just like all other languages, tools are being developed for interlanguage translation from sign language to text; however, the complexity of the sign language makes it difficult to create a computer vision model. The focus of this study is to tackle the problem with a vision-based approach using deep learning. In this work, an 8-layer Convolutional Neural Network (CNN) aimed at sign language recognition has been proposed. The model extracts feature from input images using convolutional filters before passing them to dense layers. A validation accuracy of 99.34% has been achieved on a self-created dataset comprising of 52 classes. The model has also been tested on a publicly available dataset to evaluate its efficacy.