Research On Static Gesture Recognition Technology Based On Monocular Vision

With the continuous breakthrough and development of the level of science and technology,human-computer interaction has become the focus of intelligent life.Compared with other man-machine interaction,gesture recognition technology has the characteristics of nature,convenience,economy and good user experience.At the same time,it also has good market application value and scientific research value.In recent years,gesture recognition technology has attracted more and more attention in computer interaction technology,but how to achieve a general solution,the ordinary camera,good robustness to the change of the external environment,as well as unmarked real-time gesture recognition system is not a good solution yet.In view of this problem,this paper studies the correlation algorithms of unmarked static gesture recognition from two aspects of hand detection and gesture recognition.The main work and research content of this paper are as follows:(1)The original image data is acquired by single camera,and then the image data is preprocessed,which is mainly related to image noise and light change.(2)In the study of hand gesture detection,according to the characteristics held by the hand,the method of multi feature fusion is used to realize hand detection,which is about 60% higher than that of only Haar-like.According to the distribution characteristics of human skin color,HSV color space is used to express different skin color due to the influence of external illumination on skin color change.Because the hand is a non-rigid body,it is easy to fuse the background information when collecting gestures.In this case,the background of gestures is removed when processing gesture images using Haar-like features.In addition,we use the fast AdaBoost classifier.It improves the speed of Haar-like feature classification.(3)In the research of static gesture recognition,this paper studies the gesture recognition method based on the SIFT features of the skin color.SIFT feature is based on the gray image,the other information outside the gray space inside especially interference background extraction hand features area is relatively large,by increasing the skin color information helps to distinguish between background and hand,can solve the background disturbance problem to some extent.At the same time,the HOG feature is added to enhance the robustness of the scale and rotation of the opponent,and it is of great significance to improve the gesture recognition.