Research And Implementation Of Gesture Recognition Vital Algorithm In Human-computer Interaction

With the rapid development of human-computer interaction technology, human-computer interaction technology in gesture recognition has become an important topic in the field of computer vision, the gesture interaction is a human-computer interaction in a very important way. Therefore, the gesture recognition technology has received extensive attention from all sectors of society.Imagine a scenario, when you sit in front of the TV, you need not remote control, you only use the hand to turn on the television and television channel switching. The main purpose of this paper is based on such an application. By the bare hand gesture recognition, combined with the simple algorithm, realizing the screen switching functions of the terminal in the effective and real-time ways.The gesture recognition technology is widely used, but it is similar to other target recognition that have a lot of difficulties and challenges, especially the challenges and difficulties encountered more dynamic hand gesture recognition. Gesture challenges mainly come from the unpredictable environment and gesture characteristics, such as illumination changes and gesture features similar background interference, target occlusion and so on. In order to reduce the difficulties of gesture recognition, this paper assumes that the movement gesture as rigid objects. In order to accurately, real-time segmentation of video moving gesture. This paper presents a method of combining the skin color model and gesture motion characteristics of the video motion gesture segmentation algorithm. In order to better carry out the skin color segmentation of hand gesture images, through experimental and theoretical methods, training and test the gesture picture for model selection color plane in HIS color space. In order to enhance the robustness, this paper has taken for skin color segmentation, parameter-Gauss mixture model based on pixels and in order to reduce the complexity of the whole algorithm in this paper. By down sampling method, while retaining the overall segmentation effect at the same time, reducing the complexity of Gauss mixture model, a real-time segmentation for video gesture, and then combined with motion information and region growing method to extract the feature points, so as to effectively eliminate the interference, the implementation of video gesture segmentation algorithm:Firstly, the algorithm used the GMM model in the HIS color space to extracts the GMM parameters by using the training samples, and tested the gesture video by using the model and parameters of the trained. Test algorithm mainly consists of five parts:1) get the I-frame and the J-frame from each capture video images;2)sampling the I-frame and the J-frame to get the sampling points, then the sampling points are segmented to acquire color sampling points by skin color model respectively.3)screening feature points by combining the relative motion information of the I-frame and the J-frame with region growing method;4) to obtain feature points extraction of center of gravity and the motion vector, and then determine the gesture motion direction. A large number of test results show that this algorithm is simple, effective, and real-time to switching the screen function.