Research On Static Gesture Recognition Method Based On PCA-HOG And LBP Feature Fusion

With the continuous development of the computer,people on the way of human-computer interaction put forward higher requirements.As gestures have intuitive,natural and other characteristics,so gesture recognition has become an important way of human-computer interaction.But the gesture itself has the diversity,as well as the differences in time and space and other characteristics,making gesture recognition as a highly challenging multi-disciplinary research topics.How to quickly and accurately identify the gestures of the meaning expressed,has become the focus of people's research.In this dissertation,a static gesture recognition system based on computer vision is designed and designed to recognize the recognition of six predefined static gestures.In this dissertation,several common image preprocessing methods are discussed to remove the noise of the image and enhance the quality of the image.The gradient histogram and the support vector machine are introduced respectively.Due to the complexity of gesture image background and the diversity of gestures,we choose the most powerful HOG feature of single feature.Compared with other features,HOG feature has strong robustness to the change of light and small rotation of gesture image,The HOG feature is combined with SVM as a recognition algorithm for gestures.The experimental results show that HOG combined with SVM has better classification effect on gesture recognition.However,in the classifier training,the traditional HOG feature extraction dimension is higher,there are a lot of redundant information,making the algorithm more complicated.In order to overcome this problem,an improved algorithm is proposed.The main analysis method is introduced to reduce the HOG feature to form a new PCA-HOG feature and integrate with the LBP feature to form a new PCA-HOG + LBP fusion gesture feature.This feature has both gestational edge gradient information and texture feature information,which can effectively compensate for the deficiency of single HOG feature and improve the recognition rate of gesture in occlusion.Finally,we use the gesture image in the Jochen Triesch gesture database to validate the recognition algorithm.The results show that the recognition algorithm based on PCA-HOG + LBP feature can improve the real-time performance while improving the recognition rate of gestures.Based on Open CV design and implementation of the gesture to identify the prototype system.The hardware and software platform and the system flow in the experiment are discussed,which proves that the system is feasible in practical application.