深度学习视角下的体操运动员动作识别与评估研究  

Study on Gymnast Movement Recognition and Evaluation from the Perspective of Deep Learning

作　　者：韩坤霖[1] 张煜新[1] 陈宪章 Han Kunlin;Zhang Yuxin;Chen Xianzhang(Shaanxi Energy Vocational and Technical College,Xianyang 712000,Shaanxi)

机构地区：[1]陕西能源职业技术学院,陕西咸阳712000

出　　处：《现代科学仪器》2021年第2期234-238,共5页Modern Scientific Instruments

摘　　要：为了解决肉眼观察运动员体育训练的动作到位程度和精准度误差大、效率低、智能化程度低的问题,降低人员主观因素对运动员动作评估和优化的负面影响。本文在分析深度学习经典的卷积神经网络、循环神经网络算法的优缺点的基础上,结合复杂场景、复杂动作、快速运动下对人体动作的识别、分析,提出了改进的卷积网络和优化后的神经网络混合神经网络算法。通过实验证明,该算法实现了类似人类观察运动员的训练执行情况和完成度。与其他算法的对比实验,验证了其对于运动员的动作识别具有非常高的学习性和准确性。In order to solve the problems of large errors,low efficiency and low intelligence of the manual naked eye observation of athletes'sports training,reduce the negative impact of subjective factors on the evaluation and optimization of athletes'movements.this paper analyzes the advantages and disadvantages of the classical convolution neural network and cyclic neural network algorithm,combined with the complexity For the recognition and analysis of human actions in scenes,complex movements and fast movements,an improved convolution network and an optimized neural network hybrid neural network algorithm are proposed.Experiments show that the algorithm can realize the training execution and completion degree similar to human observation of athletes.Through the comparative experiment with other algorithms,it is verified that the algorithm has the ability to recognize athletes'movements Very high learning and accuracy.

关 键 词：体操运动 动作识别 深度学习 

分 类 号：G807.01[文化科学—体育训练]