| Classroom teaching is the most common form among all teaching forms in our country.The learning state of students in classroom has an impact on students’ classroom learning and teachers’ teaching.It is not only the main factor affecting students’ learning quality,but also an important basis reflecting teachers’ teaching quality.In the traditional classroom teaching,teachers have to pay attention to the learning state of all students while teaching,which is undoubtedly a big test for the inexperienced new teachers.At the same time,in order to evaluate the teaching quality of teachers,the teaching administrators can only observe the teaching situation of students and teachers through the method of manual in-depth classroom.Such an approach not only fails to understand the real classroom situation of students and teachers in a timely and comprehensive way,but also can be influenced by personal subjective consciousness and make subjective judgments.With the vigorous development of the Internet,the Internet of Things,artificial intelligence and other high and new technologies,the field of education has also entered the era of education information.Using information technology to analyze students’ classroom learning state has become the current mainstream,but few researchers have analyzed the learning state in the real classroom environment.Based on the above reasons,this research aims to use computer vision technology to monitor the learning state of students in real classrooms in real time,so that teaching administrators can more timely and efficiently understand the degree of interest and acceptance of students in the teaching methods and content of teachers.This approach can solve the onesidedness and limitation of traditional classroom teaching administrators’ evaluation of teachers’ teaching quality.At the same time,predict the future classroom learning state,allowing teachers to timely adjust the teaching methods or adjust the classroom atmosphere when students continue to be confused,and maintain the students’ good learning state,thereby improving the teacher’s classroom teaching efficiency and the student’s classroom learning efficiency.In order to achieve the purpose of real-time monitoring of students’ learning status in class,this research constructed a database of students’ classroom images in real classroom environment according to students’ facial features in class.On the basis of this database,an improved neural network model is used to analyze the student’s classroom learning status,and the Long Short-Term Memory(LSTM)is used to predict the development trend of students’ learning state.The main research content is divided into the following parts.(1)First,by collecting video images of students in real classrooms,cutting out representative student images.At the same time,by observing students’ video images and combining with previous studies,the categories of students’ classroom learning states and the corresponding facial features of each state are summarized.Then,together with relevant professionals such as pedagogy and psychology,etc.,we annotated the learning state categories of students’ images artificially,and constructed an image database of students’ classroom learning state in real classroom environment.(2)Aiming at the problems of face head deflection,face occlusion,and poor sharpness in the real classroom image database,an improved ResNeSt network model is proposed on the basis of the existing neural network model to improve the accuracy of the recognition of students’ classroom learning state in this paper.The model introduced spatial attention mechanism on the basis of the original ResNeSt network,and made it form a complementary relationship with the channel attention thought in the original network,so that the model could extract the key features of the student images in this paper more effectively,so as to improve the accuracy and robustness of the network.(3)In order to allow teachers to adjust teaching methods and teaching contents,adjust classroom atmosphere and maintain students’ good classroom learning state when students continue to be confused,this paper uses an improved LSTM(Long and Short Term Memory Network)to analyze the data of students’ historical learning state in classroom and predict the trend of students’ learning state in the future based on these historical data.In this way,the teaching content and classroom atmosphere can be adjusted when students are constantly confused or wandering,so that students can maintain their interest and enthusiasm for the class and improve their classroom learning efficiency.The work of this study is summarized as follows: firstly,the image database of students’ learning state is constructed in the real classroom environment,and then the images of students’ learning state are identified and classified.Finally,the development trend of students’ learning state is predicted,so as to provide guidance for teachers to adjust the teaching method and content and improve students’ learning efficiency in class.In order to verify the effectiveness of the improved algorithm proposed in this paper,this paper makes a comparative experiment on the original network and the improved network in this paper on the basis of the database built by ourselves.It is verified that the proposed method can achieve better accuracy on the dataset in this paper through the commonly used algorithm evaluation criteria. |
