基于注意力机制的深度学习体育运动姿态估计技术  

Sports posture estimation technology based on deep learning of attention mechanism

作　　者：魏玉福 陈丽萍[1] WEI Yufu;CHEN Liping(School of Health,Baotou Medical College,Baotou 014030,China)

机构地区：[1]包头医学院卫生健康学院,内蒙古包头014030

出　　处：《电子设计工程》2023年第2期152-155,共4页Electronic Design Engineering

基　　金：国家自然科学基金(61662052)。

摘　　要：随着计算机技术的飞速发展,姿态估计技术在体育训练以及指导、虚拟现实应用等多个方面扮演着越来越重要的角色。文中提出了一种基于注意力机制的深度学习姿态提取技术,该技术为了更加精准地获取运动姿态信息,将注意力机制与姿态估计深度学习网络模型充分结合,实现了运动姿态的准确提取,并以一组人类运动姿态数据为实验数据,与其余两种姿态估计方法进行了实验对比,结果表明,该文提出的方法在姿态检测平均精度方面提升了1.06%,效果更佳。With the rapid development of computer technology,posture estimation technology is playing an increasingly important role in sports training,guidance,and virtual reality applications. In this article,we propose a deep learning gesture extraction technology based on the attention mechanism. In order to obtain more accurate motion posture information,the technology fully combines the attention mechanism with the posture estimation deep learning network model to achieve the accuracy of the motion posture.Extracted and used a set of human motion posture data as experimental data,and compared it with the other two posture estimation methods. The results show that the method proposed in this paper improves the average accuracy of posture detection by 1.06%,and the effect is better.

关 键 词：姿态提取 深度学习 注意力机制 体育姿态 姿态估计 

分 类 号：TN-9[电子电信]