Gesture based implementation on neural network: Bengali sign language to text conversion

Abstract

This paper presents a novel system that converts Bengali Sign language to text using an optimum system comprising of artificial neural networksand support vector model. Bengali Sign Language, has very few research papers based upon it. Therefore, for this paper we have created our own dataset. Microsoft Kinect is utilized as its input device and a complex method of bone and joint determination using contour shape, fingertip detection. Contour feature is extracted and is run through a Support Vector Machine for classification of the sign. The program has been developed in Visual Studio 2015 with a SVM wrapper.