Computer Vision-based Gesture Recognition Technology Human-Computer Interaction

With the widely application and rapid development of digital media technology,human computer interaction become more and more important in our daily life.It is hoped that human can interact with computer more naturally and efficiently.Despite of the emergency of new hardware and application,the interaction modes between human and computer are still limited to keyboard and mouse.As a result,these interactions confine the speed and naturality between human and computer.Because of the limitations of computer technology,there is still distance between human and computer.Vision based hand gesture recognition become a key method, through which people can implement a new generation human-computer interaction mode.Hand gesture is a natural and intuitive interaction mode between people.It is no need for a mid-media between human and computer communication if one use hand as the input of computer.User can define a proper hand gesture to control machines around.It is more natural using hand gesture recognition as an interface between human and computer.Research results are made based on references:(1) The introduction, background and application are discussed and new method of hand gesture interaction is proposed.The main trchniques are discussed. The Architecture of hand gesture recognition system and hand localization,tracking,segmentation and recognition are discussed.(2) Since there are disadvantages in skin color detection and motion detection ,a method of skin color combined with motion detection is proposed to locate hand. The hand is segmentated effectively in background.(3) A method based on improved CamShift algorithm is proposed to track hand. Kalman filter is used in CamShift to solve the problem of large area of skin color with low computational cost, and to predict next state. Experimental result show the effetivity of this method.(4) Segmentation is performed using adaptive thresholding method with fast computation and low computational cost. Experimental result show that the backproject image from CamShift is processed by morphology method, and output a good segmented image.(5) Proposed a geometric moment combined with contour detection method to recognize hand gestures.the moments are obtained from the segmented image by geometric moment features.contour is detected from the gray image. Directional feature of contour is obtained by histogram. Finally, the distances between images are calculated by weight of features.In this paper, with the technology of computer vision,digital image processing and pattern recognition,the computer percepts hand gestures through vision channel(camera),then act as user wants, including browse images and resize images ,realize natural human-computer interaction.