Abstract:

Gesture recognition is the growing fields of research. Being a significant part in non verbal communication hand gestures are playing vital role in our daily life. Hand Gesture recognition system provides us an innovative, natural, user friendly way of interaction with the computer which is more familiar to the human beings. By keeping in mind the similarities of human hand shape with four fingers and one thumb, this paper aims to present a hand gesture recognition on the basis segmentation of input image and Convolutional neural network based classification of hand gesture.
In this paper, we propose a method for recognizing isolated as well as continuous Kannada and Telugu alphabet gestures which is a step towards helping and educating the hearing and speech-impaired people. We have performed the classification of the gestures with convolutional neural network.

INTRODUCTION:

With the advancement of the technology, different methods of interaction with the computers have developed. More traditionally with a keyboard, mouse and thereafter joysticks, trackpads, electromechanical gloves, etc. have been used. Apart from these methods, gesture recognition has also been used and this can be considered as a more natural mode of interaction since it mimics normal conversation with a human being. Therefore, it provides a good interface for human-computer interaction (HCI). Hand gesture recognition has many applications like general computer interaction, gameplay, sign language recognition, robotic device manipulation, etc.
Gestures can be identified as a meaningful body motion which can involve hands, head, body in order to convey a meaningful information or to interact with the environment. These gestures can be static, dynamic or both [1]. In dynamic gesture recognition, it is required to identify both spatial and temporal movements and this paper proposes a methodology for dynamic gesture recognition.
In overall, gesture recognition can be divided into two broad areas and those are vision-based methods and methods require special hardware like gloves [2], armbands and body kits [3]. The second type is comparatively difficult for the user since it is required to carry required hardware when the user needs to use the system. These types of systems might be required for specialized applications like Unmanned Arial Vehicle (UAV) [4] control but not for a general computer user. On the other hand, vision-based systems use techniques like image processing, pattern recognition, image extraction, image segmentation and object detection [1]. Using vision-based approaches provides a non-obstructive interface for the user and thus created a boost in this area [5].
In this paper we have proposed Image based hand gesture detection form still image. Here we created dataset of 34 different hand sign image then performed preprocessing and segmentation method on image and then use of convolutional neural networks for classification.