Research Of Face Gender Recognition Based On CNN

The technology of gender recognition based on face image,not only can provide personal basic information for some reach fields,and also can improve the accuracy of human identity recognition,face gender recognition technology is a hotspot of research on face image.In numerous image recognition method,Convolutional Neural Network(CNN)structure has the advantage to process two-dimensional image,the recognition of image rotation and displacement deformation also has good robustness,CNN had obtained the widespread application.Research on the CNN theory and applied it to face gender recognition has theoretical significance and practical application value.This paper has great developed the research of face gender recognition based on CNN method,has finished the following works:1.Outlined the research status of Convolutional Neural Network theory and face gender recognition technology,makes a detailed review for the related theory and technology.2.By analysis of the CNN training process,in allusion to the problem of the slows convergence speed,using Particle Swarm optimization(PSO)algorithm to improve CNN,let the training parameters of CNN be PSO,and the error function be fitness function.The simulation results shows that: comparison with the normal CNN convergence rate has increased.3.Designs a kind of convolutional neural network model uses for face gender recognition called FGI(Face Gender Identification),it contains a local binary pattern(LBP)pretreatment layer,two convolutional layer,two sampling layer and a fully connected layer.The simulation results show that the comparison with LeNet-5 model identification precision has improved.4.The algorithm that PSO to improve FGI neural network model(called PSO-CNN algorithm)uses in the face gender simulation experiment based on the AR face database,comparison with the classical BP neural network recognition method?the normal CNN recognition method,and covered face image simulation experiment,verified the PSO-CNN algorithm on face genderrecognition method has fast convergence speed,high accuracy and has good robustness,proves that PSO-CNN algorithm is a effective face gender recognition method.