| 引用本文: | 徐 韬1,陈孟元2,3,程云麟1.基于 YOLOv7-SSW 模型的智慧课堂行为识别算法(J/M/D/N,J:杂志，M：书，D：论文，N：报纸).期刊名称,2026，43（1）：64-71 |
| CHEN X. Adap tive slidingmode contr ol for discrete2ti me multi2inputmulti2 out put systems[ J ]. Aut omatica, 2006, 42(6): 4272-435 |
|
| 摘要: |
| 目的 现有的学生行为识别模型存在检测精度较低且预测速度不足的问题,基于此,提出一种基于 YOLOv7-
SSW 模型的智慧课堂行为识别算法,用于提升检测精度的同时保障算法的实时性。 方法 首先,在 YOLOv7 的主干
网络引入 SE(Squeeze-and-Excitation Module)注意力机制,建立通道相关性增强算法对信息的敏感性,进而提升算
法的检测精度;其次,为了增强算法的实时性,将原始采用串行通道设计的 SPPCSPS 模块改进为并行通道设计,有
效提升了算法的检测速度;最后,引入 Wise-IoU 损失函数增强算法处理训练数据中的低质量样本能力,进一步提
升模型检测精度。 结果 所提方法在 STBD-08 数据集上的 mAP(mean Average Precision)达到 91. 9%,与 CBPHNet、YOLOv7 算法相比,分别提升了 4. 4%、5. 6%,且单帧推理时间仅 45. 1 ms。 结论 所提算法实现了学生课堂行为
的实时准确识别,为推动思政教育从知识灌输向行为素养培育模式转变提供了技术支撑,对推动教育数字化创新
发展具有重要意义。 |
| 关键词: 高等教育 思政教育 学生行为识别 YOLOv7 |
| DOI： |
| 分类号: |
| 基金项目: |
|
| Behavior Recognition Algorithm for Smart Classroom Based on YOLOv7-SSW Model |
|
XU Tao1，CHEN Mengyuan2 3，CHENG Yunlin1
|
|
1. School of Business Anhui University of Technology Ma?? anshan 243032 Anhui China
2. School of Electrical Engineering Anhui Polytechnic University Wuhu 241000 Anhui China
3. Key Laboratory of Advanced Perception and Intelligent Control of High-end Equipment Wuhu 241000 Anhui China
|
| Abstract: |
| Objective Existing student behavior recognition models suffer from low detection accuracy and insufficient
prediction speed. To address this this paper proposes a smart classroom behavior recognition algorithm based on the
YOLOv7-SSW model aiming to improve detection accuracy while ensuring real-time performance. Methods First the
Squeeze-and-Excitation SE attention mechanism was integrated into the backbone network of YOLOv7. This integration
strengthened the modeling of inter-channel correlations and improved feature sensitivity thereby improving the detection
accuracy of the algorithm. Second to enhance real-time performance the original serial-channel Spatial Pyramid Pooling
with Cross Stage Partial Connections Structure SPPCSPS module was redesigned into a parallel architecture
significantly accelerating the detection speed. Finally the Wise-IoU loss function was integrated to strengthen the
algorithm?? s capability in handling low-quality samples in training data further boosting detection accuracy. Results The proposed method achieved a mean Average Precision mAP of 91. 9% on the STBD-08 dataset representing
improvements of 4. 4% and 5. 6% over the CBPH-Net and YOLOv7 algorithms respectively. The single-frame inference
time was only 45. 1 ms. Conclusion The proposed algorithm enables real-time and accurate recognition of student
classroom behaviors providing technical support for transforming ideological and political education from knowledge
indoctrination to behavioral literacy cultivation. It holds significant importance for advancing digital innovation and
development in education. |
| Key words: higher education ideological and political education student behavior recognition YOLOv7 |
