Deep Convolutional Neural Networks-Based Sign Language Recognition System

Abstract

Deaf individuals rely heavily on sign languages. They make use of them to communicate with others. Although deaf individuals are familiar with sign language, but it is not widely understood by the general public. In this article, sign language recognition through convolutional neural network (CNN) system is developed for those who aren't familiar with sign language. American sign language letters are utilized in this work. We tried to create a translator for these letters for people who do not know sign language, and we used GoogleNet, a CNN, using the transfer learning method. Our dataset was used to train the network. The network model and network weights are recorded for the test data set once network training is finished. The accuracy of this sign language recognition system is 91.02%.