Research Of Skin Color And Gesture Recognition Method

Abstract:In recent years,with the rapid development of computer technology,the researchers are to carry out a lot of work on intelligent human computer interface of understand natural language and gesture interaction.Control methods of remote controller have become unfashionable,but tend to be more humanized control.The gesture input as the way of human-computer interaction more natural and concise.And at the same time compared to other modes of human-computer interaction,gesture input is more direct and rich.However,before by using hand gestures or people’s posture to control the machine,we must cut the correct color target from the images.But how can the image correctly cutting the target area of skin color is a very complex problem,because the skin color range is very wide,and will be affected by different lighting conditions, relax skin threshold and easy cut things like skin color in daily life.So,to design a kind of method that can overcome these uncertainties is of great theoretical significance.There are two major research topics in the thesis:(1)Used the Probability Neural Network for image segmentation,devised a method of optimal cutting threshold,in order to cut the skin color in different complex background as purpose.Join the Probabilistic Neural Network to determine the target program of skin classification,and filter out non-skin-color target like skin colors.And Experiments show that this method has better recognition effect by using eight different gesture classification.(2)By comparing the Graph Cuts (GC),Probabilistic Neural Networks(PNN) and GC+PNN three kinds of methods for cutting the skin results.Experimental results show that the combination of GC and PNN of the cut rules,not only improves the simple use of image segmentation method for cutting the success rate, but also to reduce the computation time, and it has the better recognition effec.