Recognition of Static Hand Gestures of Alphabet in Bangla Sign Language

Abstract

This paper presents a system for recognizing static hand gestures of alphabet in Bangla Sign Language (BSL). A BSL finger spelling and an alphabet gesture recognition system was designed with Artificial Neural Network (ANN) and constructed in order to translate the BSL alphabet into the corresponding printed Bangla letters. The proposed ANN is trained with features of sign alphabet using feed-forward back-propagation learning algorithm. Logarithmic sigmoid (logsig) function is chosen as transfer function. This ANN model demonstrated a good recognition performance with the mean square error values in this training function. This recognition system does not use any gloves or visual marking systems. This system only requires the images of the bare hand for the recognition. The Simulation results show that this system is able to recognize 36 selected letters of BSL alphabet with an average accuracy of 80.902%.