Static Gestures Recognition Based On Computer Vision

With the development of human-computer interaction technology, people are no longer satisfied with the traditional mouse and keyboard as an input port, an urgent need for a new type of human-computer interaction more natural pattern. Human-computer interaction models are not computer-centric to user-centric interactive mode - gesture recognition, it is now the trend. At the hospital, come to clinic patients with the general population using the same touch navigation equipment, healthy people through contact with infected devices have a chance to pathogenic bacteria. Computer vision-based gesture recognition technology a good solution to this problem. It can be operated by a simple navigation system hospitals gesture, without access to an external input device, avoiding the infection-causing bacteria.This article describes the background and application of computer vision -based gesture recognition technology, an overview of the basic theory of gesture recognition technology, gesture recognition process involves research into related technologies. Hand constructed constraint model, the use of digital gesture one to nine, to build gesture image template for gesture image recognition.Whole gesture recognition work is divided into four phases:Gesture image segmentation, image preprocessing, edge detection, feature extraction matching identification. Firstly, the comparison for each color space, RGB space and H component selection HSV color space to extract the combination of color, image segmentation gesture. Then, after binarization image filtering, threshold selection filter and Gaussian filter for image noise reduction, the use of open computing technology to remove salt and pepper noise in the image. After using the canny edge detection operator, smoothing the edges of the image after the extraction, calculated gestures edge contour.Finally, according to the outline of the characteristics of gestures, the use of Hu invariant moments of gesture contour characteristic parameters of the seven feature extraction, matching gesture recognition through the completion of the Euclidean distance.This paper studies gesture recognition applications to achieve translation, size and rotation invariant. Use OpenCV library, the validity of each module, real-time verification. Obtained through gesture recognition module performance and statistical data, experiments in the context of a simple gesture one to nine gesture recognition experiments, the recognition rate of 92%. In the complex background, gesture recognition rate of 87%.