Research On Hand Gesture Recognition Algorithm

Frequent interaction between human and computer has become a daily operation in daily life.The research on hand gesture has become one of the main research directions in the field of human computer interaction.Research on gesture recognition technology will change the traditional way of human-computer interaction,using gestures will make the human-computer interaction technology from the machine as the center gradually shifted to people-centered,make human-computer interaction more convenient,human-computer interaction becomes rich,also reduced the threshold of computer.The gesture recognition system complete expression,mainly includes four main parts,static gesture image preprocessing,image segmentation feature extraction on gestures,gestures and final gesture recognition method.The system through the camera to capture the gesture images,the image preprocessing,including transformation,smoothing,morphological operation,grayscale,binarization,contour extraction in color space,which details the commonly used color space,analysis of influence to gesture feature extraction and segmentation of color components,and weakened by the color space conversion even eliminating the influences.In this paper,the detection method based on Canny edge is introduced in detail,and the improvement is proposed based on the shortcomings of the gesture edge extraction.Gesture segmentation part is one of the key steps in gesture recognition,gesture segmentation algorithm is not high in a simple,single background of the indoor environment and outdoor environment but in complex background,there are too much interference,which makes the traditional segmentation methods cannot be gestures from the background clean out of the traditional segmentation.This paper introduces the Otsu algorithm in a single context although the effect is good,but the complex background is difficult,Otsu algorithm by dividing the improved method of gray image segmentation can make a gesture.The amount of information in the clean gesture image too much,as if the classification system input,computation time increases the recognition system and the computational complexity,so feature extraction of gesture image is needed,the use is unchanged from the image features of the geometry,unchanged from the 7 invariant distance value.We have the rotation,translation and scale invariant recognition system in order to make the classification of the input,we need through the simulation and comparison of selected in line with the conditions of the component and combined into the input vector.In the selection of identification method,the method of adaptive neural fuzzy inference system(ANFIS)is selected in this paper,which has the ability of autonomous learning and good robustness.Although this method has better identification ability but high computational complexity,we based on the gesture recognition method unchanged from the screening combined with adaptive neuro fuzzy inference system,improve the gesture recognition rate of the system,and with the BP neural network and fuzzy neural network,the average recognition rate of 95.3% shows that adaptive neuro fuzzy inference system in the effect of the recognition rate,in line with the actual criterion of high recognition rate.