<abstract xmlns="http://www.w3.org/1999/xhtml">

<p>In the context of the new curriculum, physical education has become an important part of school teaching. In this paper, the edge extraction algorithm and TV-L1 optical flow method are utilized to extract key image features in physical education classroom teaching, and the improved TE network structure extracts the skeletal data features of teachers and students in the classroom based on the TEMS module. Then, the MTHA-MDCNN deep migration model is established to fuse the three features to identify multimodal teaching behaviors in the sports classroom. Experiments show that the accuracy of this paper’s method for recognizing teaching behaviors in physical education classrooms can reach 77.63%. In the actual behavior recognition, the average values of teachers’ and students’ behavioral coverage are 32.68% and 67.77% respectively, which is in line with the optimal division of teaching time in physical education classrooms. In contrast, the students’ behavioral coverage of physical practice is low, which is not conducive to their physical and mental health development. The construction of teachers and supervision of teaching should be strengthened in schools, and students’ learning paths should be expanded. The multimodal analysis method of teaching behavior in this paper provides a reference method for targeted improvement of physical education teaching, and the proposed optimization strategy is important for the successful optimization of physical education classroom teaching in schools in the future.</p>
</abstract>

In the context of the new curriculum, physical education has become an important part of school teaching. In this paper, the edge extraction algorithm and TV-L1 optical flow method are utilized to extract key image features in physical education classroom teaching, and the improved TE network structure extracts the skeletal data features of teachers and students in the classroom based on the TEMS module. Then, the MTHA-MDCNN deep migration model is established to fuse the three features to identify multimodal teaching behaviors in the sports classroom. Experiments show that the accuracy of this paper’s method for recognizing teaching behaviors in physical education classrooms can reach 77.63%. In the actual behavior recognition, the average values of teachers’ and students’ behavioral coverage are 32.68% and 67.77% respectively, which is in line with the optimal division of teaching time in physical education classrooms. In contrast, the students’ behavioral coverage of physical practice is low, which is not conducive to their physical and mental health development. The construction of teachers and supervision of teaching should be strengthened in schools, and students’ learning paths should be expanded. The multimodal analysis method of teaching behavior in this paper provides a reference method for targeted improvement of physical education teaching, and the proposed optimization strategy is important for the successful optimization of physical education classroom teaching in schools in the future.

Multimodal Analysis and Optimisation Strategy of Teaching Behaviour in Physical Education Classroom

Applied Mathematics and Nonlinear Sciences

This work is licensed under the Creative Commons Attribution 4.0 International License.

In the context of the new curriculum, physical education has become an important part of school teaching. In this paper, the edge extraction algorithm and...

Multimodal Analysis and Optimisation Strategy of Teaching Behaviour in Physical Education Classroom

Publicado en línea: 02 jul 2024

Páginas: -

Recibido: 20 feb 2024

Aceptado: 13 may 2024

DOI: https://doi.org/10.2478/amns-2024-1506

Palabras clave
Feature extraction, TV-L1 optical flow method, Skeletal data features, MTHA-MDCNN, Teaching behavior

© 2024 Guannan Liu, published by Sciendo

This work is licensed under the Creative Commons Attribution 4.0 International License.

Multimodal Analysis and Optimisation Strategy of Teaching Behaviour in Physical Education Classroom

Publicado en línea: 02 jul 2024

Páginas: -

Recibido: 20 feb 2024

Aceptado: 13 may 2024

DOI: https://doi.org/10.2478/amns-2024-1506

Palabras claveFeature extraction, TV-L1 optical flow method, Skeletal data features, MTHA-MDCNN, Teaching behavior

© 2024 Guannan Liu, published by Sciendo

This work is licensed under the Creative Commons Attribution 4.0 International License.

Palabras clave
Feature extraction, TV-L1 optical flow method, Skeletal data features, MTHA-MDCNN, Teaching behavior