Gesture Recognition Method Based On Computer Vision Research

With the development of science and technology,people-centered human-computer interaction has become the inevitable trend of the times.Gestures,in addition to the language in daily life is another way of communication,is one of the important bearing objects express ideas.Gesture to today,but also become the important media for human-machine interaction,hand gesture recognition based on computer vision,can achieve non-contact human-computer interaction and meet the new demand of the development of the times.In this paper,the problem of hand gesture recognition based on computer vision is studied,and ASL gesture library is used to preprocessing of gesture images,feature extraction and recognition methods.The main contents of this paper are as follows:First,this paper introduces the preprocessing of gesture image.This part contains the binarization of gesture images,gesture image smoothing,gesture image edge processing and gesture image morphological processing and other aspects of the content.Preprocessing of gesture images enables gesture recognition to be better and faster,and is a very important part of gesture recognition.In the subsequent gesture recognition,combined with different feature extraction and classification methods,this paper also tried different image preprocessing methods,for the subsequent gesture recognition to provide a more effective support.Second,this paper analyzes the gesture recognition method based on feature extraction and support vector machine.This part focuses on gesture feature extraction.It mainly analyzes and compares gesture recognition using different features such as SIFT features,PCA features and HOG features.Through the feature extraction methods such as PCA,HOG and SIFT,we extract some of the features which are more explanatory,and then use these features to represent the image,and then classify them through the SVM classifier.The experimental results show that the HOG feature is more suitable for hand gesture recognition,and the gesture recognition based on HOG feature is better than the other two methods,and the recognition rate is the highest.Third,this paper analyzes the gesture recognition method based on convolution neural network.This section introduces convolution neural networks to adaptively learn gesture characteristics through depth networks.This part analyzes the influence of each parameter in the CNN network on the gesture recognition,and in view of the network training time is long,the problem such as a large demand for samples,this part uses an already trained network to initialize the convolution neural network,which speeds up the network convergence speed.Experimental results on the ASL data set show that.Compared with the traditional feature extraction algorithms,the gesture recognition based on convolution network not only has better robustness,but also gets better recognition results.