American sign language classification using deep learning

Abstract

Image classification is a process that incorporates analysing and extracting useful information from an image. It addresses a wide range of real-world issues and has applications in the fields of artificial intelligence, robotics, biomedical imaging, motion recognition, among many others. In this paper, we have utilised support vector machine (SVM), decision trees (DT), k-nearest neighbour (kNN), convolutional neural networks (CNN), VGG-16, ResNet-50, MobileNet-V2, and DenseNet-201 on American Sign Language dataset. This paper describes a system that uses deep learning and machine learning to recognise gestures in images and assign an English alphabet corresponding to the gesture. The results will be useful for classification of American sign language as the necessary comparison metrics and performance of all these models are studied and documented in this paper.