Research On Emotion Recognition Method And Application For Online Education

With the rapid development of Internet and multimedia technology,how to realize online education intelligently has become a hot issue in the education field.In the online education,the abundant learning resources can be individualy provided by the prominent open communication environment and the teaching & learning mode is independent of time and space.However,the existing online education systems are not flexible enough to adjust learning contents for different learners in real time.Moreover,the learners’ emotional information reflects their own learning needs,so it is important and necessary to improve the flexibility of the online education systems by mining the emotional state of the learners so as to recommend some personalized learning contents.Machine learning provides an effective way for data-driven emotion analysis.Oriented the online education system,this paper focuses on constructing the learner’s emotional model,and developing its corresponding machine learning based emotion recognition method.The main work is as follows:Firstly,the facial emotion recognition method is studied based on cost sensitive convolution neural network.In order to solve the problem of poor practicality of the model caused by ignoring the cost of loss resulting from misclassification,a cost-perceived convolutional neural network based facial emotion recognition method is proposed.This method integrates the loss cost caused by the error classification into the model training process,reduces the loss cost of misclassification and improves the practicability of the emotional recognition model.Secondly,the text emotional recognition method based on Word2Vec-LSTM is studied.For the problem ignoring the importance of text interactive information,a text emotion recognition method based on Word2 Vec and LSTM is designed.This method uses the Word2 Vec model to extract the deep semantic features of the texts,uses the LSTM model to extract its time sequence feature,and obtains its corresponding emotional type by the softmax classifier.Finally,how to apply emotion recognition model into online education systems is explored.On the basis of the facial emotion recognition algorithm realization,the facial and textual emotions are combined together to comprehensively assess the learner’s emotional state,and the corresponding evaluation results are used into online education systems to adaptively recommend the contents and types of learning resource for learners so as to meet their individualized needs.