Research On Hand Gesture Recognition Technology Based On Deep Learning Bayes

With the rapid development of science,gesture recognition,as an important research direction,has the advantages of flexibility and wide application.From relying on external devices at the early stage,it gradually developed to use computer vision algorithm.However,manual design is needed to extract gesture features in the algorithm,which leads to too complicated specific implementation process.And there are still some difficulties in the process of gesture recognition,such as the distortion and stretching of gesture image,or the influence of too many parameters in the neural network on the recognition results.All of these factors pose challenges in recognizing gestures from images.Therefore,in view of the above problems,this paper proposes an improved YOLOV3(You Only Look Once Version3)gesture recognition network model in the field of deep learning in terms of target detection technology.The main research work is as follows.First present for general network vulnerable to external factors,lead to problems such as poor test results,use a space transformation of YOLOV3 network structure was improved,and according to this article adopted by the gesture image data size and constitute the basis of design of transformation network positioning in the network space structure and the related parameters,enhance the invariance of the network;Then K-means clustering method is used to optimize the prior box parameters in the network,and CIOU detection method is used to optimize the judgment process of threshold value.Then,an improved gesture recognition classifier was constructed by combining principal component analysis(PCA)with Naive Bayes classification method to reduce dimension and classify features extracted from the network.Finally,this paper adopts the method of simulation experiment of the proposed method to verify the validity of the improved,using the publicly available data sets and contained I make gestures of the image data sets,to improve the model for model training and testing,the mean average detection precision of the improved model can reach 96.73%,compared with other models,this paper puts forward the improved model has obvious promotion on gesture recognition accuracy,can improve the gesture recognition accuracy,the effectiveness of the proposed algorithm,with good research value and effectiveness.