Study On Learning State Based On Facial Features

With the continuous development of Internet technology and multimedia technology,online learning has become increasingly mature.With its unique advantages in time and space,online education can meet the learning requirements of different groups.It has made a unique contribution to promoting the balance of education in China,making up for the shortcomings of education,and opening the era of universal learning.Its existence also enriches the existing Educational model.However,compared with traditional classrooms,online learning cannot supervise the learner's learning in real time,which leads to the disadvantages that the learning effect cannot be guaranteed.Therefore,detecting the learner's learning status is very valuable for improving the online teaching system and improving the learning quality of the students.At the same time,the test results are fed back to teachers,which plays an active role in optimizing classroom settings for teachers and comprehensively evaluating students' learning situation.In this article,we use computer vision,deep learning and other technologies to conduct research the learning state from two aspects: attention state research and emotion recognition.The main research contents of this article are as follows:1?The traditional detection method cannot detect the distracted state of the learner's sight when the focus is on the screen.For this problem,this paper proposes an attention detection method based on eye movement analysis.This method uses computer vision related technologies to Extract the eye movement features from the real-time captured pictures,analyze the sequence of eye movement features using the recurrent neural network,learn the eye movement laws,and realize the detection of the learner's attention state.2?Based on facial expression recognition,the study of emotions is studied,and seven emotions such as surprise,happiness,and fear are classified into three emotion states:positive,neutral,and negative,and emotion state is detected through the recognition of expressions.Traditional convolutional neural networks have shown great capabilities in processing computer vision-related tasks,but there are also problems that can not use different levels of features and feature redundancy.For this kind of problems,in order to achieve better detection effect,this paper proposes a multi-feature fusion multi-scale Represents a learning network.This method can effectively integrate different levels offeatures through cross-connection operations to improve the learning ability of the model,while using global average pooling to reduce the amount of parameters.Constructing Octave convolution blocks further reduces the spatial redundancy of the model and enhances robustness.This method can effectively improve the recognition accuracy,while greatly reducing the training parameters.The learner's learning picture were captured in real time by using the computer's own camera.Then analyze the eye movement characteristics and its attention,and judge the emotional state.The experiment result shows that the accuracy of the two methods proposed in this paper is significantly improved compared with the same type of algorithm,which effectively solves the problems of attention detection and facial emotional state recognition when the learner's eyes are on the screen.