Sign Language Translator Using Deep Learning Techniques

Sign language is mainly used by hearing impaired people for communication. And it has been an active research topic for the past 2 decades. Although sign language seems to be familiar in recent times, it's still quite a challenge for non-signers to communicate with signers leading to the widening of communication gap between them. Deep learning models are widely being used for motion and gesture recognition. With advancement in deep learning techniques and computer vision, this topic is gaining more attention. The main aim of the proposed model in this paper is to implement a sign language translator using deep learning techniques like Convolutional Neural Network (CNN) and help to bridge the gap between signers and non-signers. The model recognizes all the numbers[0-9], alphabets[a-z] and custom user-defined unigram symbols with better accuracy both in real time and for static sign image classification. A web application is built with the trained model to make it accessible to everyone. The front end of the web application is built with the help of a python library called gradio for simple and user friendly interface.