Research On Facial Expression Recognition And Online Education Emotion Evaluation Based On Deep Learning

Today,with the rapid development of emerging technologies such as artificial intelligence,5G Internet of Things and cloud technology,deep learning algorithms are becoming more and more mature,and facial expression recognition based on deep learning has also become a research hotspot in the field of machine learning.The application of facial expression recognition technology The field is also becoming wider.In recent years,facial expression recognition has been used for research in the field of "artificial intelligence education".In the context of the new crown epidemic,it is also an inevitable development trend to promote the application of artificial intelligence in online education.Therefore,this paper studies the facial expression recognition technology based on deep learning,and proposes an online education emotion evaluation scheme based on facial expression recognition.The main research work completed in this paper is as follows:（1）Aiming at the low accuracy of facial expression recognition today,a facial expression recognition method based on an improved deep residual network model is proposed.First,by improving the original deep residual network Res Net18 by deepening the network,an improved deep residual network C-Res Net18 is proposed,and then a self-attention weighting module is introduced to construct a weighted deep residual network model C1-Res Net18,The self-attention weighting module outputs a corresponding weight for each face image,and the weight is used to weight the feature vector of the corresponding expression image.Experiments are carried out to verify that the facial expression recognition accuracy rate reaches 98.89% and 87.13%,respectively,which is significantly improved than the facial expression recognition rate of the original deep residual network Res Net18 model and several other network models,which proves that this method can It can effectively improve the accuracy of facial expression recognition and has certain application value.（2）Aiming at the large number of parameters of convolutional neural network and the low recognition accuracy of large-scale facial expression dataset,a facial expression recognition method based on improved Efficient Net network model was proposed.First,an improved deep neural network model C-Efficient Net is proposed,which uses the Efficient Net-B5 network to extract facial expression features,then uses the self-attention mechanism for weight learning,and then uses the multi-layer perceptron（Multi-layer Perception）to learn.,MLP)for sample feature fitting and expression classification,and the Gaussian Error Linear Unit（Gelu）function was introduced as the activation function.Finally,three public datasets,CK+,RAF-DB and Fer2013,were used to model the proposed model C-Efficient Net conducted experiments to verify that the accuracy of facial expression recognition reached 98.91%,87.74% and 70.63%,respectively.The experimental results show that the accuracy of facial expression recognition is better than that of other current models.（3）A research on emotional evaluation of online teaching of facial expression recognition was carried out.On the one hand,an improved feature fusion network framework C-Efficient Net2 based on Efficient Net and Res Net18 is proposed.Efficient Net-B5 and C-Res Net18 proposed in the previous（2）are used as the backbone network to extract image features in multiple dimensions and introduce feature fusion.To improve the recognition accuracy,a new activation function is also proposed:exponential-logarithmic linear function.The proposed model C-Efficient Net2 is experimentally verified by using the public datasets CK+,RAF-DB and Fer2013,and its facial expression recognition accuracy reaches 98.94%,88.04% and 72.32%,respectively.The experimental results show that the improved network model can effectively improve The accuracy of facial expression recognition.On the other hand,an online teaching emotion evaluation scheme based on facial expression recognition is proposed.The real online classroom video is decoded to obtain video frames,and then a face detection module is designed to obtain a series of facial expressions of teachers and students in classroom teaching.Pictures,use the improved deep neural network model to identify the expressions of teachers and students,evaluate the online teaching emotion,and evaluate the online teaching effect from one dimension.