Research On The Recognizing And Matching And Retrieval Method Of Hand Gestures Based On Computer Vision

As a novel man-machine interaction technology,the hand gesture interaction is supirior in some areas,such as virtual reality,sign language interpretation,remote control and so on, because of its simplicity,intuition and ease of use.Therefore,it obtains more and more attentions.Although it is still a hot spot of technical research,the hand gesture interaction begins to be applied in the real life gradually,which you may find in some latest consumer electronic products.It represents a new trend of interacton in the future and indicates the broad application perspect of the hand gesture interaction.There are three key problems must be solved before the implementation of the hand gesture interaction:the detection and recognition of hand gestures,the shape matching of hand gestures and the retrieval of hand gesture instructions.In the thesis,efforts of exploration and research have been made around the three key points and a method of the recognizing and matching and retrieval method of hand gestures based on computer vision has been proposed,which aims to dectect the position of human hands and recognize the shape of the hand in a monocular video sequence, then match it with the pre-defined hand gesture models based on the shape context algorithm,thus realize the goal of hand gesture instructions recognition.In order to extend and maintain the hand gesture instructions set conveniently,store the feature parameters of the hand gesture models to save the time spent on the recalculation and improve the retrieval efficiency,a solution about the storage,management and retrieving of the hand instructions set using databases has been given in the thesis.The contributions of the thesis are listed as follows:1) The features of a variety of previous gesture model have been summarized.Conbimed with the charactoristics of the method of shape context,a definition of a new gesture model is proposed.The model is characterized by using the shape context to convert the standard hand image to the log-polar histogram,as an important property of the gesture model.2) With the combination of the skin color detection and differential motion analysis,good effect on the segmentation of the hand region from video sequences can be achieved.Compared with the previous methods, this method overcomes the limitations of using a single method and has strong applicability and stability.3) The shape context matching method is proposed to be applied in the matching of the human hand gestures. Compared with previous shape matching methods used in gesture interaction research,the shape context method uses the two-dimensional invariance of log-polar transformation,and overcomes the shortcoming of the traditional shape matching algorithm.It is proved that this shape matching method is stabile and effective,and it is applicable to human hand gestures shape matching.4) This thesis researches on the content-based image retrieval technology.Combined with the gesture model proposed in this paper,the thesis models data structure for gesture information and uses database technology to create,store and manage gesture instruction set.In this way,it is easy to maintain and manage and query the hand gesture set through the application programs.This paper revolves around how to apply the gesture recognition technology in the practical environment as the core objectives and has meaningful exploration.In this paper,this computer vision-based gesture recognition matching retrieval method is an effective and practical method,which mainly uses the shape context method,not only to get good results in the shape matching of gestures,but also to provide a new solution for the gesture modeling and searching.