Research Projects

The deaf and dumb population often struggle to communicate effectively with health workers due to the lack of support for sign language. This project proposes a real-time sign language to text conversion system to bridge this language barrier. The project aims to design an end-to-end framework to support communication between the deaf and dumb and health workers. Most sign language recognition systems in literature only consider a small set of sign language data and cannot accurately identify gestures under different lighting conditions and complex backgrounds. There are no standard datasets available for Indian sign language gestures, and there is no defined system for a complete end-to-end framework. The main hypothesis is to design a robust model that can efficiently identify sign language under different lighting conditions and complex backgrounds. NIT silchar has proposed various sign language recognition algorithms, which will be tested in real-time environments. The model will be obtained using deep learning techniques like Convolutional Neural Network, Transformers, Visual Transformers, and GAN. The prototype system will be compact and affordable, accessible to a wide class of medical practitioners.