Research On Facial Expression Recognition Based On Hybrid Domain Attention Mechanism

As the most natural and direct way for human expression of inner emotions,human facial expression plays a vital role in interpersonal communication and social life.Compared to linguistic expressions,expressions often express more precise information than words.In today's rapid development of artificial intelligence and computer vision,people are more eager to understand the human emotions,and the role of facial expression recognition is particularly critical.Facial expression recognition has a wide range of application scenarios,such as human-computer interaction,intelligent transportation,criminal investigation,etc.With the continuous development of "Internet +" and "AI+",facial expression recognition will have more applications in future.The early facial expression recognition method is a two-step method based on feature extraction and classification of traditional machine learning.Firstly,the feature extraction algorithm is used to extract the features in the image,and then the classifier is selected for feature classification.The traditional recognition method has a fast recognition speed and does not require a large data set,but the recognition accuracy is relatively low.For the traditional method,the design of the feature requires some prior knowledge,and the quality of the feature design will directly affect the accuracy of the model,which hinders the improvement of recognition accuracy.Today,deep learning has become a hot trend in the field of artificial intelligence.Different from the traditional machine learning method,deep learning is an "end-toend" learning method,that is,the feature extraction and classification are put into the same model.Deep learning uses deep nonlinear neural networks to gradually integrate the underlying features of data into high-level features.This method can better describe the essential information of data.Based on the above analysis,this paper discusses the application of deep learning in the field of computer vision,analyzes several commonly used convolutional neural network models,and proposes a simple convolutional neural network model for facial expression recognition.Firstly,this paper designs a basic model based on the hierarchical structure of convolutional neural network,and uses various optimization methods such as weight decay and batch normalization to optimize the model to prevent over-fitting,etc.,based on the characteristics of small amount of data in the expression recognition data set.Second,use the attention mechanism to optimize the model to improve recognition accuracy.The attention mechanism is inspired by the principles of human vision,with the goal of selecting the most important information from the many tasks for the current task.The addition of the attention mechanism can strengthen the weight of the key features,making the model more focused on the features useful for the classification of expressions during the training process.This article uses the attention mechanism of the hybrid domain,focusing on both channel attention and spatial attention.Finally,the method proposed in this paper is analyzed experimentally.The experimental results show that the attention mechanism can effectively improve the accuracy of expression recognition.