A Research Of Gesture Recognition Based On Feature Extraction

The gesture is a very important means of communication between people in the study of human-computer interaction is also very large practical. With the diversification of the interaction between the development of computer technology, computer, gesture recognition technology involves an interdisciplinary field of image processing, pattern recognition, computer vision, coupled with the diversity of gestures, multi ambiguity, and time and the spatial variability of characteristics, gesture recognition will naturally become an attractive research topic. The gesture recognition process can be divided into gesture modeling, hand data acquisition, identify the five stages of the hand data preprocessing, feature extraction of hand, hand characteristics. In this paper, the static gesture recognition, mainly through modeling analysis may be effective in the characteristic parameters of the target gesture, and then analysis based on the outline of the image feature parameter extraction method, and finally by the gesture recognition algorithm based on template matching to extract the characteristics The parameters of the experiment, comparison of their validity.Firstly, the comparative analysis of three-dimensional model of the gesture-based modeling and performance-based gesture modeling, pending identification of the target sample for this study design the gesture of a performance-based model-staffing the Department of Tortoise model, it has the advantage fingers and the palm to distinguish a good description of the basic features of the hand, to facilitate the analysis of characteristic parameters. Tortoise model staffing department to get a number of potentially effective characteristic parameters, including the number of fingers outstretched in the gesture, the ratio of the number of defects of hand contour and gesture contour area and perimeter. In addition, the image of the moment invariant features as image recognition of the important parameters in this study analysis of seven Hu moment invariants as characteristic parameters.The focus of this article pretreatment part of general adaptive thresholding method and the gesture segmentation segmentation method based on Otsu’s threshold effect, the experiments show that the use of gesture segmentation method based on Otsu threshold segmentation than the general adaptive threshold segmentation method.This article feature parameter extraction part by the Ministry Tortoise staffing model analysis to work out the outline of calculation of the palm of your hand, the radius of the palm of your hand, palm center of gravity, the ratio of the defects in the number of fingers the number and profile area and perimeter, especially through the convex polygon verification four rules seek the palm of your hand contour and contour based on the palm of your hand, fingers the number of the calculation method. In addition, this study Hu invariants, to achieve a calculation based on the outline image of Hu moments invariants, seven Hu moments invariants.Finally, a detailed analysis of the calculation results of the10feature parameters, all the characteristic parameters are divided into three groups, three groups were compared the effectiveness of the characteristic parameters, the design of a gesture recognition system based on template matching, experimental results show that extraction ten characteristic parameters for gesture recognition is valid.