A Study On Real-time Gesture Hci Based On Binocular Vision

Computer vision-based gesture human-computer interaction is a non-contact human-computer interaction,It allows the user to get rid of a particular input device,It makes information exchange between computers is even more direct and natural.Vision-based gesture recognition and tracking used single or multiple cameras and combined with computer vision techniques to understand the gesture,Then use the video stream to calculate the trajectory of hand gesture.In this paper,first of all,we detect the hand gestures adaptivly through double images acquisition,then used to track and locatefingertips.Hand Gesture tracking is used to forecast the general area of hand gesture in the next frame.The coordinates of the fingertips can describe the type of gesture and the three-dimensional reconstruction of fingertip.The main content of this paper and innovation are as follows:(1) This paper presents a hardware devices design of strictly synchronous double images collection,for synchronous problems when collect images.The device can be synchronized control at the pixel level.The device is ready for follow-up digital image processing and Three-dimensional reconstruction(2) This paper presents an adaptive detect and tracking method after analysis and comparison of a variety of classic gesture detection and tracking algorithm.This algorithm combines the color detector and gaussian mixture model detector to locate the gesture area and compute it’s rate.which improves the weights update method in gaussian mixture model.It is able to reconstruct the true background rather than change stationary target into background.It also can overcome the shortcomings of "ghost".In this paper,the tracking method based on particle filter used to predict the gesture to reduce the time-consuming.This paper also presents a method of combine with multi-geometric characteristics to locate the fingertips.(3) In this paper, the above methods is used to calculate the fingertips’ coordinate, Then used the epipolar correction method to make the same of fingertips’ y-coordinate.It can achieve the stereo matching of fingertips.In the end of this paper,we reconstruct th fingertips’ three-dimensional coordinates based on parallax and camera parameters.