Sign Language Recognition using Hand Gestures

Abstract

Sign language is one form of language for communication, but since most people do not know, this research work has come up with a method using neural networks for fingerspelling based on American Sign Language. In our method, a snapshot of a hand gesture of a real time video is taken and is passed through a filter known as a Gaussian blur filter, which is then passed through CNN to predict the class of the hand gestures. The proposed method provides an accuracy of 95.8% for the 26 letters of the alphabet, thus it acts as an assistant for speech impaired people.