Abstract
Vision based gesture recognition technologies based on structural shape descriptors and contour analysis methods have been described in this paper. Gesture have been the basic mode of communication since ancient times, until vocal communication was developed. Skin color detection technique for hand segmentation based on YCbCr methods has proved effective enough in real environment. Haar based classifier for face detection is used to remove one of the largest contour besides hand. Median filtering along with morphological operations applied here alleviates the effects of noise to a great extend without losing the boundary information in the image. The binary image formed after this only consists of two largest skin colored contours, mostly one or two hand contour image. Over this resultant contour, structural analysis based on contour shape along with convex hull and convexity defect formation together with geometrical analysis based on angle between convexity defect point and that of hand centroid helps in determining the number of fingertips. Based on number of finger-tips, the gestures can be classified into various gestures. Various applications in real time hand gesture recognition include American Sign Language (ASL) recognition, human computer interaction (HCI), robotics, real time traffic signal control and remotely controlling devices such as TVs, air conditioners, fan using hand gestures.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call