Scholarlink

<p>The effectiveness of Physical Education (PE) teaching plays a vital role in encouraging students’ physical development, engagement, and lifelong health performances. Conventional evaluation methods frequently suffer from objectivity and struggle to reflect the many components of PE instructions. This research intends to establish a comprehensive assessment framework for PE teaching effectiveness that uses Multimodal Machine Learning (MML) to evaluate complicated, multi-source educational data. The multimodal dataset captures instructional delivery (video/audio), physical movement patterns (via wearable sensors and motion capture), and student perceptions. To ensure data quality and consistency, two distinct preprocessing methods are employed: noise reduction by wavelet denoising to clean sensor signals and z-score normalization to control data across various modalities and verify balanced input distributions. The Histogram of Oriented Gradients (HOG) is used for feature extraction, capturing fine-grained movements and posture characteristics from video sources, and generating the spatial representation of physical activities. The essence of the assessment framework is the Enhanced Marine Predators-driven Explainable Boosting Machine (EMP-EBM), a hybrid system that combines the EMP’s global optimization strengths with the EBM’s transparent and interpretable structure. The proposed EMP-EBM model achieved better performance, with an accuracy of 98.07%, a precision of 97.85%, a recall of 98.31%, and an F1-score of 98.07%, demonstrating its strong capacity to identify and assess efficient PE teaching sessions by utilizing multimodal inputs. By distributing both precision and explainability, this model advances the use of Artificial Intelligence (AI) in education, making the calculation process more objective, adaptive, and aligned with the multifaceted goals of PE.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A comprehensive evaluation system for physical education teaching effectiveness supported by multimodal machine learning

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年12月28日

查看全文
  • Yanqing Zhang

摘要

The effectiveness of Physical Education (PE) teaching plays a vital role in encouraging students’ physical development, engagement, and lifelong health performances. Conventional evaluation methods frequently suffer from objectivity and struggle to reflect the many components of PE instructions. This research intends to establish a comprehensive assessment framework for PE teaching effectiveness that uses Multimodal Machine Learning (MML) to evaluate complicated, multi-source educational data. The multimodal dataset captures instructional delivery (video/audio), physical movement patterns (via wearable sensors and motion capture), and student perceptions. To ensure data quality and consistency, two distinct preprocessing methods are employed: noise reduction by wavelet denoising to clean sensor signals and z-score normalization to control data across various modalities and verify balanced input distributions. The Histogram of Oriented Gradients (HOG) is used for feature extraction, capturing fine-grained movements and posture characteristics from video sources, and generating the spatial representation of physical activities. The essence of the assessment framework is the Enhanced Marine Predators-driven Explainable Boosting Machine (EMP-EBM), a hybrid system that combines the EMP’s global optimization strengths with the EBM’s transparent and interpretable structure. The proposed EMP-EBM model achieved better performance, with an accuracy of 98.07%, a precision of 97.85%, a recall of 98.31%, and an F1-score of 98.07%, demonstrating its strong capacity to identify and assess efficient PE teaching sessions by utilizing multimodal inputs. By distributing both precision and explainability, this model advances the use of Artificial Intelligence (AI) in education, making the calculation process more objective, adaptive, and aligned with the multifaceted goals of PE.