Age Classification Methods Based On Deep Convolutional Network And HMAX Model

Face as one of the most significant biometric features,contains a wealth of personal information.This information can be used not only for identity authentication and identification,but also in the application of face age classification.It can be used in age and distribution characteristics of human-computer interaction and business intelligence applications.Age classification research has important scientific significance and practical value.In the extraction of face age information,the traditional neural networks and SVM classification methods are usually affected by complex non-linear factors,such as individual genetic differences,living environment,health status,racial differences and other factors.To overcome the impact of these factors,this paper uses HMAX model to extract features inspired by visual cortex on the preprocessed facial images and performs SVM on these features to get reasonable age classification results.Then,based on the deep convolution neural network.After local perception,weight sharing and translation invariant,a series of convolution,pooling,full connection,normalization layer and other methods,the pater gets better age classification results compared with the previous HMAX based method.The main contributions of this paper include:(1)A method of face classification based on HMAX model.Firstly,the active shape model is used to extract the facial features.To reduce the effects of translation,scale and rotation,we normalize the face image.Secondly,we use HMAX model to extract C1-S features.Thirdly,we use SVM to classify the facial features.(2)The face age classification based on deep convolution neural network.Considering that the depth convolution neural network usually need large amount of training sets,we enlarge the number of training facial images.We then propose a deep convolutional neural network structure which has good classify performance.We choose the right activation function,reasonable normalized layer number and Dropout retentionprobability which have great influence on the network performance through experiments.After parameter tuning,the convolutional neural network we proposed could get better classification performances.The face age classification algorithm based on deep convolutional neural network and HMAX model proposed in this paper have achieved better accuracy in FG-NET face data set.It has the ability to complete different age classification tasks and can be effectively used in many fields such as security,human-computer interaction,entertainment audio and so on.