Research On Expression Feature Extraction And Recognition Algorithm

With the continuous improvement of artificial intelligence and computer processing technology,as a branch of face recognition,facial expression recognition technology is gradually emerging.Researchers have tried to make computers understand human emotions,and let computers read facial expressions become a hot topic of research.The traditional facial expression recognition method is based on the characteristics of the researchers'manual design,which has low requirements on computer performance,but the recognition rate is limited.The neural network algorithm relies on its own distributed feature representation,after training enough facial expression images.The recognition accuracy is significantly higher than other traditional methods,but it requires higher computing power.This paper studies the current facial expression feature extraction and recognition algorithms.The main work includes the following:1.Introduce the main components of facial expression recognition,describe and summarize the existing feature extraction and recognition algorithms,analyze and compare the advantages and disadvantages of various feature extraction and recognition algorithms,and introduce the common facial expression recognition.The data set,and the image preprocessing process,including face detection and normalization.The feature extraction algorithms of RILPQ(rotation invariant phase quantization)and CSLBP(improved local binary mode)are studied.Combining the advantages of CSLBP and RILPQ,an expression recognition based on fusion feature extraction is proposed.A CSLBP is used to describe the texture features of the image.RILPQ is used to describe the different regions.Then the two sets of features are separately encoded to generate feature maps.Then the block and statistical histograms are used to obtain the feature vectors of the expressions.Using the histogram combination method,The SVM classifier is used to classify facial expressions,and the expression recognition of single feature extraction is improved,which improves the recognition accuracy.The expression recognition based on CNN(convolution neural network)is studied.By designing a reasonable neural network model,adding nonlinear representation in the input layer,and adjusting the optimal network parameter training network model,the data set division ratio is determined through experiments.Realized facial expression recognition.The experimental results under the JAFFE data set show that the algorithm can obtain higher recognition rate than the traditional algorithm.Finally,the face expression recognition system based on CNN model is developed.The face expression recognition GUI interface is designed by Pycharm and Tkinter.The system interface displays the expression recognition result in real time and verifies the effectiveness of the expression recognition system.The expression is recognized in real time,which verifies the generalization ability of the expression recognition system.