Vision-based Gesture Recognition And Its Application Research

With the development of computer technology,the importance of the interaction between man and machine is becoming more and more important.The initial human-computer interaction was done via the keyboard input command line,followed by a mouse,a touch screen,and so on.The use of gesture as the input of human-computer interaction is more natural and convenient.The vision-based gesture recognition has a wide application prospect,which can be used for safe driving,business office,sign language recognition,entertainment facilities,etc.Meanwhile,it brings many convenience to people's daily life.These application prospects also make many domestic and foreign scholars devote themselves to this field Research.In this paper,based on the vision gesture recognition,following work were done:Firstly,in the aspect of hand gesture segmentation,the video frames acquired by the camera are mixed with the histogram equalization to improve the contrast of the image.Based on the skin color method,the YCrCb color space is used to segment the static gesture.A geometry method is used to remove the arm part,and finally obtain the palm part of the image.Secondly,in the aspect of static gesture recognition,we use the Hu moment feature of gesture image as the gesture feature,and use BP neural network to train it,and get the static hand gesture classifier.9 kinds of custom gestures are trained and recognized,with more than 95%recognition rate which is at a high level in the industry.Then,in the aspect of dynamic gesture recognition,this paper improves the existing TLD tracking algorithm in combination with the previous static gesture recognition results,and realizes effective and real-time tracking of the gesture.Six common gestures are trained using the HMM with an 88.7%recognition rate,and the recognition effect is good.Finally,based on the above research of gesture recognition,this paper develops two programs to study the application of gesture recognition,combining two specific application backgrounds.The first application for the "finger-guessing game",using the dual cameras to identify two users of static gestures,and compare and output contrast results;the second application for the gesture control of the music player,using the gesture to control music volume and music switching and other functions.