YANG Lei,LI Yingxiang,ZHANG Hongbo.Improvement and Application of Physical Fitness Test based on Deep Learning[J].Journal of Chengdu University of Information Technology,2022,37(05):538-543.[doi:10.16836/j.cnki.jcuit.2022.05.009]
基于深度学习的体能测试计数算法改进与应用
- Title:
- Improvement and Application of Physical Fitness Test based on Deep Learning
- 文章编号:
- 2096-1618(2022)05-0538-06
- 关键词:
- 体能测试; lightweight-openpose; 模型改进; 运动识别; 运动计数
- Keywords:
- physical fitness test; lightweight-openoose; model improvement; motion recognition; motion count
- 分类号:
- TP183
- 文献标志码:
- A
- 摘要:
- 为改变体能测试中人工计数的传统方式，提出一种基于深度神经网络的体能测试计数算法。该算法以运动视频作为输入，采用轻量级的姿态估计网络(lightweight-openpose)对人体关节点坐标进行检测，改进原模型的关节漏检和无效计算问题并新增了人物追踪模块; 使用图像分类网络进行运动类型预测，采用批量正则化和迁移学习方式强化网络在自制4类常见体能测试数据集上的训练; 制定4类体能测试运动计数标准，结合姿态估计和运动识别结果进行有效运动判断并计数。实验表明，姿态估计模型相较于原模型运行速度提升17.6%，人物追踪错误率低至3.2%; 运动识别模型准确率提升5%～8%，达到94.84%; 运动计数算法平均准确度在4类体能测试运动中达到95%。
- Abstract:
- In order to change the traditional way of manual counting in physical fitness test, a physical fitness test counting algorithm based on deep neural network is proposed. The algorithm takes the motion video as the input, uses the lightweight-openpose network to detect the coordinates of human joints, improves the joint missing detection and invalid calculation of the original model, and adds a human tracking module. The image classification network is used to predict the types of sports, and the batch normalization and transfer learning are used to strengthen the training of the network on the self-made four kinds of common physical fitness test data sets. Four kinds of sports counting standards for physical fitness test are formulated, and valid test judgment and counting are carried out combined with the results of pose estimation and motion recognition. Experiments show that the running speed of the pose estimation model is improved by 17.6% compared with the original model, and the error rate of human tracking is as low as 3.2%; The accuracy of motion recognition model is improved by 5%-8%, reaching 94.84%; The average accuracy of the motion counting algorithm reached 95% in the four types of physical fitness test.
参考文献/References:
[1] 满蔚仕,朱宗耀,张志禹,等.采用同步挤压小波变换的人体运动姿态分析[J].西安交通大学学报,2017,51(12):8-13.
[2] 张永强.基于Hu不变矩特征优化的人体运动姿态识别算法[J].计算机科学,2014,41(3):306-309.
[3] Gao L,Zhang G,Yu B,et al.Wearable human motion posture capture and medical health monitoring based on wireless sensor networks[J].Measurement,2020,166(4):108252-108263.
[4] Zhao L,Chen W.Detection and recognition of human body posture in motion based on sensor technology[J].IEEJ Transactions on Electrical and Electronic Engineering,2020,15(5):766-770.
[5] Thar M C,Winn K Z N,Funabiki N.A Proposal of Yoga Pose Assessment Method Using Pose Detection for Self-Learning[C].2019 International Conference on Advanced Information Technologies(ICAIT).IEEE,2019:137-142.
[6] Osokin D.Real-time 2D Multi-Person Pose Estimation on CPU:Lightweight OpenPose[J/OL].https://arxiv.org/abs/1811.12004,2018,11,29.
[7] Wu Z,Shen C,Hengel A.Wider or Deeper:Revisiting the ResNet Model for Visual Recognition[J].Pattern Recognition,2016,90(2019):119-133.
[8] Fang H S,Xie S,Tai Y W,et al.RMPE: Regional Multi-person Pose Estimation[C].2017 IEEE International Conference on Computer Vision(ICCV).IEEE,2017:2334-2343.
[9] Zhe C,Simon T,Wei S E,et al.Realtime Multi-person 2D Pose Estimation Using Part Affinity Fields[C].IEEE Conference on Computer Vision and Pattern Recognition(CVPR).IEEE,2017:1302-1310.
[10] 闫航,陈刚,佟瑶,等.基于姿态估计与GRU网络的人体康复动作识别[J].计算机工程,2021,47(1):12-20.
[11] 苏超,王国中.基于改进OpenPose的学生行为识别研究[J].计算机应用研究,2021,38(10):3183-3188.
[12] Sandler M,Howard A,Zhu M,et al.MobileNetV2:Inverted Residuals and Linear Bottlenecks[C].2018 IEEE Conference on Computer Vision and Pattern Recognition(CVPR).IEEE,2018:4510-4520.
[13] Ioffe S,Szegedy C.Batch normalization:Accelerating deep network training by reducing internal covariate shift[C].International conference on machine learning.PMLR,2015:448-456.
[14] Priya D T,Udayan J D.Transfer learning techniques for emotion classification on visual features of images in the deep learning network[J].International Journal of Speech Technology,2020,23(8)361-372.
[15] Krizhevsky A,Sutskever I,Hinton GE. Imagenet classification with deep convolutional neural networks[J].Advances in neural information processing systems.2012,25(4):1097-1105.
备注/Memo
收稿日期:2022-01-05
基金项目:四川省科技计划重点研发资助项目(2021YFG0149)