Published Paper Details:

People can interact and exchange ideas and emotions through communication. The social contacts of the deaf community are hindered by multiple factors. The people converse with each other using sign language. A technology can translate sign languages into a form that is comprehensible in order to communicate with ordinary people. Developing a real- time text-to-Indian Sign Language (ISL) translation system is the aim of this project. For the most part, manual labor is used. In this paper, we describe a convolutional neural network-based deep learning method for classifying signs. We initially construct a classifier model using the numerical signs and the Python-based Keras convolutional neural network implementation. In phase two, a second real-time system was used to use skin segmentation to detect the Region of Interest in the frame that displays the bounding box. To forecast the sign, the segmented region is fed into the classifier model. The accuracy rating of the system for the same subject is 99.56% in poor light and 97.26% in high light. It was observed that the classifier improved with varying image capture angles and backgrounds. The RGB camera system is the main emphasis of our strategy.