Indian Sign Language Generation from Live Audio or Text for Tamil
Bandi Rupendra Reddy, Daka Chandra Rup, Rohith Mathi, Meena Belwal
Abstract
Communication is the key behind the whole human evolution, and there would be nothing possible if the stream for communicating our thoughts to one another were cut off. Language is a significant factor when it comes to conveying one's ideas, thoughts, and information to other people. Sign language is used to communicate hard-of-hearing people's opinions. Every country has different sign languages. In this paper, we have used ISL (Indian Sign Language). To ease communication between ordinary people and deaf/dumb people, we propose the design and implementation of a model that translates a live voice, audio recordings, or text of a native Indian regional language (Tamil) to text and then further matches it to sign language animations from the video animation dataset. The speech is converted to text using two deep learning models LSTM(Long Short Term Memory), Bi-LSTM, and Google API. Then the text is transformed into a sign using ISL (Indian Sign Language) dataset. The proposed models achieved 45%, 65%, and 95% accuracy for LSTM, Bi-LSTM, and Google API, respectively.