Multi-angle Facial Expression Recognition And Its Application On Based Improved VGGNet

Facial expression is one of the important ways for people to communicate information between people and it can provide a more comprehensive understanding of people's inner world.Therefore,facial expression recognition has become one of the research hotspots in the field of face recognition and it has been widely concerned by researchers at home and abroad.At present,facial expression recognition has been widely used in the fields of medicine,transportation and image analysis.Therefore,it is of great significance and value to recognize and analyze facial expressions.However,the features that selected by traditional facial expression recognition methods in the extraction of facial expression features do not have good representational ability,and the features need to be extracted manually according to people's experience.All these problems have a great impact on the recognition effect of the model,resulting in poor generalization ability of the model.With the maturity of deep learning method,the application of convolutional neural network in image processing is more and more extensive,a new research method for facial expression recognition has been developed.In view of the above problems,the specific is as follows:(1)The convolutional neural network is used to extract features of facial expressions,It has the ability to automatically select and learn features of facial expressions,and combines feature extraction and expression classification into one method,which avoids the shortcoming of insufficient feature representation ability extracted by traditional facial expression recognition methods in the feature extraction stage.VGGNet(Visual Geometry Group Network)has a continuous structure of 3×3 small convolution kernel,which enhances the nonlinear expression ability of the model and can extract higher level abstract features of expressions.Therefore,VGGNet is used to extract features of expressions.After extracting facial expression features,the extracted facial expression features were visualized and analysed.(2)To improve the VGGNet,remove the last fully conneted layer from VGGNet,design a 4 layer of the connection to the extraction of the expression characteristics of neural network training,the training of the expression by Softmax layer classification,join the Dropout layer to reduce the number and avoid over fitting phenomenon,introduce Batch normalized layer(Batch Normalization,BN)avoid network depth deepening training situation.(3)A facial expression recognition system is designed to verify the performance of the algorithm.Based on the expression recognition model trained by the above algorithm,a facial expression recognition system is developed by using PyQt to write the interface.The user makes an expression recognition system for recognition.According to the recognition results,the system returns a joke or funny picture to improve the user's current emotional state.The method designed in this paper uses Tensorflow deep learning framework to conduct experiments on RaFD,CK+ and JAFFE expression databases.Compared with other methods,the accuracy is improved to a certain extent,and the recognition accuracy reaches over 90%.The experimental results show that VGGNet can extract deeper expression features and has better effect on expression recognition.