Abstract
Recent advancements in the field of computer vision with the help of deepneural networks have led us to explore and develop many existing challengesthat were once unattended due to the lack of necessary technologies. HandSign/Gesture Recognition is one of the significant areas where the deep neuralnetwork is making a substantial impact. In the last few years, a large numberof researches has been conducted to recognize hand signs and hand gestures,which we aim to extend to our mother-tongue, Bangla (also known as Bengali).The primary goal of our work is to make an automated tool to aid the people whoare unable to speak. We developed a system that automatically detects hand signbased digits and speaks out the result in Bangla language. According to thereport of the World Health Organization (WHO), 15% of people in the world livewith some kind of disabilities. Among them, individuals with communicationimpairment such as speech disabilities experience substantial barrier in socialinteraction. The proposed system can be invaluable to mitigate such a barrier.The core of the system is built with a deep learning model which is based onconvolutional neural networks (CNN). The model classifies hand sign baseddigits with 92% accuracy over validation data which ensures it a highlytrustworthy system. Upon classification of the digits, the resulting output isfed to the text to speech engine and the translator unit eventually whichgenerates audio output in Bangla language. A web application to demonstrate ourtool is available at http://bit.ly/signdigits2banglaspeech.