Research On Facial Expression Recognition Method Based On Convolutional Neural Network

With the rapid development of artificial intelligence and machine vision technology,a wave of the development in the field of image processing has been set off,and deep learning algorithms have gradually attracted people's attention.Facial expression recognition is an important part of human-computer interaction,and deep learning algorithm is used to realize the recognition and classification of facial expression recognition has become a hot issue for scholars.The paper mainly focuses on the identification and classification of static facial expressions.Firstly,the Adaboos algorithm is used to roughly cut the facial expression image,and then the gradient integral projection and double threshold binarization are used to locate the human eye of the facial expression image to achieve accurate cropping of the facial image.Bilinear interpolation and improved equalization algorithms are used to scale normalize and grayscale normalize for the cropped images,and a facial expression image of uniform size and uniform gray scale is obtained.Considering the three aspects of the characteristics of neuron,learning rules and topology of network,a neural network architecture for recognizing and classifying facial expressions is constructed.The fixed weight Gabor wavelet is used to construct the convolutional layer,the support vector machine algorithm is used to construct the fully connected layer,the matching growth rule is used to determine the hierarchical structure of the convolutional neural network,and the back propagation algorithm is used to train the parameters of the entire convolutional neural network.Finally,an experimentally determined convolutional neural network structure suitable for facial expression recognition and classification is obtained.Fisher's linear discriminant method is used to improve the principal component analysis method for the problem of excessive dimension of Gabor wavelet.,and it is used to reduce the dimension of the facial images,which effectively solving the problems of facial image has too many dimensions and long recognition time.The improved principal component analysis method combined with support vector machine algorithm and convolutional neural network algorithm respectively carried out experiments on facial expression recognition,and compared with the traditional facial expression recognition structure,validating the accuracy and effectiveness of convolutional neural network in facial expression recognition.In addition,the facial facial recognition recognition classification system and a GUI graphical user interface for human-computer interaction were designed.