竞技武术套路中难度动作智能识别方法
Intelligent Recognition Method of Difficult Movements in Competitive Martial Arts Routine
陈威 1葛士顺1
作者信息
- 1. 淮南师范学院 体育学院,安徽 淮南 232001
- 折叠
摘要
为了提高竞技武术套路中难度动作智能识别的准确性,以竞技武术套路中运动员动作的原始骨架数据建立运动员有向时空骨架图.在时空图卷积神经网络内,引入图注意力机制与通道注意力机制,建立基于注意力的时空图卷积神经网络.在该网络内输入运动员有向时空骨架图,输出难度动作智能识别结果.研究结果表明,所提识别方法可以准确提取难度动作的特征,对竞技武术套路中难度动作的识别精度较高.
Abstract
In order to improve the accuracy of intelligent recognition of difficult movements in competitive martial arts routines,the directed space-time skeleton diagram of athletes is established based on the original skeleton data of athletes'movements in competitive martial arts routines.In the spatiotemporal graph convolutional neural network,the graph attention mechanism and channel attention mechanism are introduced to establish the spatiotemporal graph convolutional neural network based on attention.The spatial and temporal framework of athletes is input into the network,and the intelligent recognition result of the difficulty action is output.The research results show that the proposed recognition method can accurately extract the features of difficult movements,and the recognition accuracy of difficult movements in competitive martial arts routines is high,which is helpful for athletes to identify and improve difficult movements more accurately in training.
关键词
竞技武术套路/难度动作/智能识别Key words
competitive wushu routine/difficult movements/intelligent recognition引用本文复制引用
基金项目
安徽省高校人文社会科学研究重点项目(SK2019A0567)
出版年
2024
NETL
NSTL
万方数据