Research On Head Posture And Facial Expression Recognition Technology For Online Teaching Evaluation

During this pandemic,online learning and online education have become the mainstream of teaching forms which becomes a good solution to the problem of not being able to take classes offline.Because it is online teaching,it is difficult for teachers to supervise students effectively.it becomes the bottleneck of online teaching evaluation that information cannot be fed back in time whether students have been online all the time,whether they have listened carefully,whether they have comprehended and whether they are confused.it becomes the bottleneck of online teaching evaluation that information cannot be fed back in time whether students have been online all the time,whether they have listened carefully,whether they have comprehended and whether they are confused.This paper aims to provide an effective visual detection and recognition system for online teaching evaluation by using deep learning technology.The research contents of this paper can be listed as follows:(1)We take the multi-task convolutional neural network(MTCNN)as our face detection network.However,it has a huge impact on the accuracy and real-time performance of video face detection that the movement of students and the deflection of the face,the complex structure of the MTCNN network,and the amount of computation,so this paper proposes a fast face detection method by joint particle filter and MTCNN.we propose a partitioned random particle optimization resampling method.(2)In this paper,the online students are authenticated through face recognition.This paper use face recognition based subspace method where we firstly use the local binary pattern(LBP)to extract the Local features of the detected faces,then use principal component analysis(PAC)to reduce the dimension to obtain the feature vectors representing the faces of the students.we identify students by Euclidean distance.(3)In this paper,the head posture of students is used as an auxiliary criterion to assess the student’s mental state.We adopt a head pose estimation method based on support vector machine.We firstly use our face detection algorithm to extract the key points of the face and get euler angles value by key points.Finally we use support vector machine(SVM)for head pose classification.(4)Aiming at how to reduce the influence of expression unrelated features in facial expression recognition task,a facial expression synthesis method based on generative network is proposed.We build a DLGAN(Disentanglement Learning GAN)network.And in order to improve the ability to learn expression details of DLGAN,this paper introduces the self-attention module into the structure of DLGAN.Because the expression duration is short,the range of motion is small,and the number of expression samples is small,we propose an facial expression classification algorithm based on two-stream 3D neural network to enrich feature input.(5)Based on the technologies of face detection,face recognition,head posture estimation,expression recognition and other technologies in this paper,we build an online teaching effect evaluation system to achieve real-time monitoring of students through the camera of electronic devices.The system analysis students’ emotional states online,estimate their head posture and detect their behaviors such as opening and closing eyes,yawning,etc.Finally,the comprehensive evaluation results of student status are given according to all test results.