CNN based Gesture Recognition for ISL

Abstract

Abstract

Gesture recognition has always been at a peak when it comes to computer vision and human-computer interaction (HCI). Hand gestures can be used for interaction with systems like managing UAVs, medical devices, videogames, etc. In this paper, the main focus is on handicapped people, who are physically unable to speak or listen. With the help of computer interaction, gestures will be recognized. As, the same gesture by another people differs, recognizing it is the main task. The Convolution Neural Network (CNN) is used, to the extraction of features and gesture recognition, RGB dynamic gesture recognition method is used, the average accuracy received is around 97.44%, which is higher comparatively. The dataset used is the original dataset of Indian Sign Language (ISL), for learning features and the gestures of the images. The method sets the feature concatenation layer of Canny Edge and RGB images in the CNN structure, to give better performance of gesture recognition.

Keywords: Convolution Neural Network, Indian Sign Language, hand gesture recognition, Human-Computer Interaction.