摘要
为了科学评价线上瑜伽教学中学生的动作姿态,把基于深度信息的人机交互技术融合进体育课堂教学,用以解决瑜伽线上教学的相关难题。笔者在人体姿态检测的基础上基于多模态思想设计了瑜伽姿态动作评分模型。利用RGB摄像头采集瑜伽的姿态动作图,即普通的RGB彩色图像,利用骨骼提取模型从RGB图像中提取骨骼数据。将RGB图像数据和骨骼数据输入联合模型中,联合模型将输出瑜伽动作的类别以及这种动作的评分,并对提出方法的有效性进行了对比实验分析和验证。对比实验结果表明:多模态数据在多模态联合模型中表现出较大的优势,集合了RGB数据和骨骼数据的优势,可以快速准确地完成瑜伽动作的分类和评价。
In order to scientifically evaluate the action posture of students in online yoga teaching,it is important to integrate human-computer interaction technology based on in-depth information into physical education classroom teaching to solve related problems in yoga online teaching.This paper is designed based on multi-modal thinking and human posture detection.The yoga posture action scoring model uses the RGB camera to collect the yoga posture action pictured,namely the ordinary RGB color image,uses the bone extraction model to extract the bone data from the RGB image,and enters the RGB image data and the bone data into the joint model.The joint model will output the types and scores of yoga actions of such actions,and analyze and verify the effectiveness of the proposed method.The results of the comparative experiment show that the multi-modal data show greater advantage in the multi-modal joint model,combining the advantages of RGB data and bone data,which can quickly and accurately complete the classification and evaluation of yoga actions.
作者
万益
WAN Yi(Department of Physical Education,Nanjing Forestry University,Nanjing,Jiangsu 210037,China)
出处
《体育研究与教育》
2021年第4期90-96,共7页
Sports Research and Education
基金
江苏省教育厅高校哲学社会科学一般课题(2019SJA0122)
江苏省体质健康促进研究中心课题(2019B007)
南京林业大学高等教育研究课题一般课题。
