Hand Gesture Recognition for Bangla Sign Language Using Deep Convolution Neural Network

Abstract

Around the world, deaf and dumb people are sufferers of all kinds of activities due to a lack of proper sign language interpreters. Our research paper proposes a new hand gesture recognition framework toward Bangla sign language to eliminate the significant communication gap between deaf and non-sign language users. The hand was detected practicing HSV and YC <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">b</sub> C <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">r</sub> color space. In total thirty-seven (37) characters (8 vowels and 29 consonants) are recognized by deep convolution neural networks. We take 37 classes for 37 alphabets from Bangla sign language. Our framework also aided to gesture recognition system by a new dataset for the Bangla sign language. Our dataset consists of 3219 images from six different people. This new dataset facilitates us to gain an accuracy of 99.22%.