Classification and Recognition of Indian Sign Language Using Deep Learning Approaches
Author : Shivansh Kulshrestha
DPS Vasant Kunj, New Delhi, Delhi, India
Hand signs are considered to be the most effective way of interaction among humans, having a vast number of applications. Since these hand gestures are known to be the natural means of communication, they are usually used for interaction worldwide by impaired people who are unable to speak. Considering this, it is observed that more than one per cent of the Indian population is suffering from this speaking impaired-ness. This can turn out to be the main factor that will significantly impact the individuals to integrate and join a framework that will comprehend the Indian Sign Languages. Various approaches have been proposed in the literature for the classification and feature extraction of sign language; however, the majority of them are based on Machine learning (ML) approaches. In this research paper, the deep learning (DL) method has been proposed for recognising and classifying the Indian Sign Language using the Convolutional Neural Networking algorithm. All of the alphabets belonging to the simple backgrounds have been taken into consideration, gathering data from more than 100 subjects in various lighting conditions. Some observations have been made to see the effect of optimization techniques and activation functions. The proposed technique has succeeded in acquiring, classifying and recognizing the static hand signs of 26 English alphabets from A-Z with approximately 4000 images. Through this, the algorithm could achieve more than 99% accuracy on the validation dataset.