SMART GESTURE USING REAL TIME OBJECT TRACKING

Abstract

Gesture can be used to interact with the computer without any physical contact. The use of keyboard and mouse can be minimized. Gesture can be of various types. One such type is movement of hand in a particular posture. To detect these type of gestures first it must be verified that the hand is present in frame and is present in the required posture. The first one is achieved by creating a mask of the frame considering the skin color range in the HSV color space. The later part involves shape matching with some template shape. The shape matching involves computing of central moments between the mask and the template shape. The hand posture defines the start and end of gesture. All the movement of hand between start and end of gesture is tracked and gesture is recognized from the tracked data. For the purpose of recognition, Convolution Neural Network is used. An application is built on recognition. Once a gesture is recognized an event will be triggered.