%0 Journal Article
%T 基于机器学习的人体动作识别实用性研究<br>Research on the Practicability of Human Motion Recognition Based on Machine Learning
%A 严浩宸
%A 徐爽
%A 张驰
%A 孙天凯
%J Journal of Image and Signal Processing
%P 158-168
%@ 2325-6745
%D 2023
%I Hans Publishing
%R 10.12677/JISP.2023.122016
%X 当前，线上线下混合式教学成为了教育的新趋势，就体育课程的线上教学而言，对学生的上课情况进行智能识别一直是各大高校研究的热点问题。体育课程不同于其他学科，线上体育课的开展具有形式单一，考核难度大的特点。后疫情时代下的体育课程仍然具有线上线下共同授课的挑战。本文就如何有效地应对线上体育动作识别这一问题，研究并实现了一种基于机器学习的人体动作识别的方法，利用计算机视觉对运动图像和视频进行采集，采用基于Kinect的人体动作识别技术进行动作规范性的检测，根据AdaBoost算法进行面部专注性检测，实时分析学生在运动中是否专注，动作是否到位等问题。结果证明：相比于传统的线上体育课教学质量评价方法，此方法既能拥有准确的识别效果，也能够减少评判的时间，对提高体育教师的工作效率有着很好的推动作用。<br />
At present, online and offline hybrid teaching has become a new trend in education. As far as online teaching of physical education courses is concerned, intelligent recognition of students’ classes has been a hot issue explored by universities. Physical education courses are different from other disciplines. The development of online physical education courses has the characteristics of single form and high difficulty in assessment. The physical education curriculum in the post- epidemic era still has the challenge of online and offline co-teaching. This paper studies and im- plements a method of human motion recognition based on machine learning on how to effectively recognize online sports motion. It uses computer vision to collect motion images and videos, uses Kinect-based human motion recognition technology to conduct normative detection of motion, uses AdaBoost algorithm to detect facial focus, and analyzes whether students are focused in motion in real time, Whether the action is in place. The result shows that compared with the traditional online PE teaching quality evaluation method, this method can not only have accurate recognition effect, but also reduce the time of evaluation, which has a very good role in promoting the work efficiency of PE teachers.
%K 机器学习，计算机视觉，面部专注度检测，AdaBoost算法，Kinect<br>Machine Learning
%K Computer Vision
%K Face Concentration Detection
%K Adaboost Algorithm
%K Kinect
%U http://www.hanspub.org/journal/PaperInformation.aspx?PaperID=64838