Research on Students’ Classroom Behavior Recognition Based on Pose Information Extraction and Local Feature Segmentation

DOI

10.2991/aebmr.k.220502.041How to use a DOI?

Keywords

OPENPOSE posture detection; HPRNN; LFRCNN; generalization ability

Abstract

Based on the extraction of human posture information, students’ classroom behavior will be identified after local feature segmentation. Because it is challenging to collect classroom behavior samples and the school students are numerous, the existing methods are difficult to obtain good generalization ability. This paper defines six classroom behaviors of “looking at the blackboard,” “looking around,” “sleeping,” “playing mobile phone,” “taking notes” and “reading”, and uses OPENPOSE posture detection network to extract the pose information of middle school students in the image, and then identifies the head pose and the surrounding environment of hands through HPRNN and LFRCNN to obtain the student classroom behavior. Experimental verification shows that this method can identify multiple students’ behaviors in the same network under the condition of ensuring recognition accuracy, which effectively alleviates the problem that neural network is difficult to train due to insufficient sample size, and avoids the decrease of network generalization ability caused by students’ different clothing and posture to a certain extent.

Copyright

© 2022 The Authors. Published by Atlantis Press International B.V.

Open Access

This is an open access article distributed under the CC BY-NC 4.0 license.

Download article (PDF)

TY  - CONF
AU  - Chenyi Cong
PY  - 2022
DA  - 2022/05/16
TI  - Research on Students’ Classroom Behavior Recognition Based on Pose Information Extraction and Local Feature Segmentation
BT  - Proceedings of the 2022 International Conference on Urban Planning and Regional Economy（UPRE 2022）
PB  - Atlantis Press
SP  - 225
EP  - 230
SN  - 2352-5428
UR  - https://doi.org/10.2991/aebmr.k.220502.041
DO  - 10.2991/aebmr.k.220502.041
ID  - Cong2022
ER  -