Rotation Independent Digit Recognition in Sign Language

Abstract

Usually deaf people use sign language to communicate with others. A large number of people including children have disabling hearing loss. In sign language, to indicate signs, hand gestures are used, however, hands can be rotated in any angle. There are few published research work to recognize digits in sign language. In this paper, we have proposed a 10 layers convolutional neural network model using residual learning to detect digits in sign language of any angle. We have collected 700 sign digit images from which we have prepared 7000 rotated images, rotating each image in arbitrary rotation 10 times. Using our proposed model, we have got 97.28% accuracy for 7000 rotated images. Our Proposed method improves accuracy by 10.57% than the latest research work.