The Research On Facial Expression Recognition Algorithm Based On Convolutional Neural Network

In recent years,facial expression recognition has been used in many fields,such as game interaction,auxiliary medicine,criminal trials,and intelligent transportation.However,in practical applications,the collected facial expression images often have complex background characteristics,and the model training speed is too slow,resulting in low expression recognition rate and poor robustness,which is difficult to meet actual needs.As a clustering algorithm,K-Means algorithm can train the K-means clustering central data set that meets the characteristics of the data set and has good initial values,and extracts the feature as the initial value of the convolution kernel of convolutional neural network,which can solve the problem of random initialization of the convolution kernel to some extent.In this paper under the framework of convolutional neural network,combined with K-Means clustering idea,a facial expression recognition algorithm based on convolutional neural network is proposed.The main research work of the paper is as follows:1.In the process of expression recognition,if the convolutional neural network(CNN)has problems such as unreasonable setting of layers and too many parameters,the expression recognition rate of CNN is low.This paper improves on the basis of the classical convolutional neural network AlexNet structure.Firstly,the network structure of CNN is adjusted,which mainly includes the number of the convolution layers and the downsampling layers,and the adjustment of the activation function and parameter optimization method of the network,it is expected to improve the nonlinear expression ability of CNN.Then,the convolutional neural network self-contained classifier Softmax is replaced with a multi-class SVM classifier,which is expected to improve the classification ability of the model to some extent.From the performance aspect of the model,the new structural algorithm model is compared with the Alex algorithm model.At the same time,the new structural algorithm model are compared with the convolutional neural network model with different classifiers in the simulation experiment.The simulation results show that compared with the unimproved AlexNet structure,the improved structure can improve the expression recognition rate to some extent.In terms of model robustness,this structure also has certain advantages compared to other convolutional neural network structures.2.Aiming at the problem that the convolutional neural network model training speed is too slow and the face expression recognition rate is not high in complex background,the K-Means clustering idea is introduced,and a facial expression recognition algorithm based on K-Means clustering idea and CNN is proposed.Firstly,the network structure of the newly proposed convolutional neural network is taken as the overall framework.Secondly,through theoretical derivation and experimental analysis,the optimized K-Means model is designed and applied to the convolutional layer of the convolutional neural network,and the convolution kernel with initial value is obtained to extract the expression image feature of the training set and the test set.Simulation experiments show that the proposed algorithm increases the feature extraction ability of the model to some extent and reduces the training time of the model.