Bangladeshi Sign Language Recognition using fingertip position

Abstract

Sign language is the only means of communication for deaf and dump people which uses manual communication and body language to convey meaning. For any sign language, an interpreter is essential to communicate with deaf and dump people. To enhance interaction with community, Sign Language Recognition (SLR) is a growing field of research now a days. The task of SLR is language specific and a number of prominent works are available for few major languages. On the hand, the works are very few for Bangladeshi Sign Language (BSL) although Bangla is a major language and Bangladesh has a large community of deaf and dump people. In this study a BSL recognition scheme has been investigated based on fingertip position. The method considered relative tip positions of five fingers in two dimension space and position vectors are used to train artificial neural network (ANN) for recognition purpose. The method seems efficient with respect to ANN training with pixel values of image as of previous studies. The proposed method has been tested on a prepared data set of 518 images of 37 signs and achieved 99% recognition rate. The proposed method is found better than exiting BSL recognition methods.