Design and Development of Indian Sign Language Character Recognition System

Abstract

The purpose of this study is to look into the challenges involved in categorizing Indian Sign Language (ISL) characters. While a lot of research has been done in the related field of American Sign Language (ASL), not as much has been done with ISL. Lack of standard datasets, obscured traits, and variance in language with geography are the key barriers that have hindered much ISL research. Our study aims to progress this field by collecting a dataset from a deaf school and applying various feature extraction techniques to extract useful information, which is then input into a range of supervised learning algorithms. Our current results for each approach include four fold cross-validation. What sets our work apart from earlier research is that the validation set in our four-fold cross-validation contains photographs of people who are not the same people as those in the training set. Hand gestures and signs are used by those with speech impairments to communicate. Understanding what they're trying to say is challenging for the average person. Though extremely uncommon, there are many systems that convert data to Hindi. Therefore, it is imperative to implement a system that enables the general public to understand and interpret all signals, gestures, and communications. It will close the communication gap that exists between normal people and those who have speech difficulty. The two primary research approaches centered on human-computer interaction are sign language recognition and learning. Multiple sensors are required in order for data flow to be understood in sign language. This research paper focuses on the development of a Hindi-language training tool that can detect images and interpret what someone else is trying to say to persons who have speech impairments. Keywords: Indian Sign Language (ISL), American Sign Language (ASL), Feature Extraction, Supervised learning, Sign Language, etc.