Monocular Visual Servo Gesture Recognition And Human-computer Interaction Research

As a new generation of human-computer interaction method, monocular hand gesture recognition based on computer vision has important theoretical research value and practical application prospect. This article researches on the main process of monocular hand gesture recognition, including gesture image pretreatment, feature extraction and hand gesture recognition. Besides, achieves the recognition of five common hand gestures which were input from ordinary camera under Microsoft Visual Studio 2008, and builds a simple gesture interactive system, thus validate the feasibility of the algorithm. At last, develops a teaching prototype system with practical application based on gesture interactive system, to illustrate the realistic significance of monocular hand gesture recognition.First, collect gesture images from video stream obtained from camera, and build hand gesture model library after pretreatment. In the pretreatment phase of hand gesture image, in order to obtain good segmentation result, this article adds artificial markers for the hand gesture. In the phase of extracting hand gesture feature, this article uses the distance distribution feature of gestures binary image as the robust feature of hand gesture recognition. It can describe the hand gestures'overall shape information through describing the hand gesture image's shape features by distance distribution feature. Meanwhile, using this method in gesture recognition has the features of invariance in translation and scaling. Because choose image gravity as the benchmarks, and the targeted areas of all images were divided into several distance areas with same number, the position of the target in image, and the scaling sizes won't affect the distance distribution features. Using this method in recognition, hand gestures don't change when it changes in revolution, translation and size, and it can develop the robustness of hand gesture recognition.This article mainly focuses on the improvement of algorithm based on monocular robust hand gesture recognition and its application in human-computer interaction. The originalities of this article mainly display in the following three aspects:firstly, put forward using pixel distance distribution method to extract the robust features of gesture areas, and achieve in natural gesture recognition. Compared with general methods, the biggest characteristic of this method is its invariant in rotation. Real-time gesture recognition can get rid of interruption in revolving images. Thus develop the robustness of hand gesture recognition. This article presents the method of adding artificial markers, removing the interruption of similar skin color region. This is an originality of this article. Using pixel distance distribution method in hand gesture recognition, needs high requirement in skin-color segmentation accuracy. In complex background, if interruption of similar skin color region exists, the distance distribution features of hand gesture may change considerably. It can directly affect the result of the recognition. This article solve this difficult issue successfully by adding artificial markers. Thirdly, the human-computer interaction system prototype based on hand gesture recognition has simple teaching function, and its application fields will be expanded.