基于骨骼关键点的散打动作识别与评价方法  

A Method for Recognition and Evaluation of Sanda Movements Based on Bone Key Points

作　　者：张冰冰 赵柏山[1] ZHANG Bingbing;ZHAO Baishan(School of Information Science and Engineering,Shenyang University of Technology,Shenyang 110870,China)

机构地区：[1]沈阳工业大学信息科学与工程学院,沈阳110870

出　　处：《微处理机》2023年第6期36-39,共4页Microprocessors

摘　　要：为提高散打运动辅助打分的公平公正性,提出一种基于骨骼关键点的散打动作识别与评价方法。首先采集十类散打动作构建实验数据集,获取视频关键帧,对图像进行去噪处理。其次通过基于YOLOv5s-CBAM目标检测的HRNet-DSC-CBAM人体姿态估计方法提取人体骨骼关键点坐标。最后利用ST-GCN动作识别方法进行动作识别,并利用DTW动态时间规整算法完成动作评价。实验结果表明,该方法在自制数据集的10类散打动作中表现出良好的识别效果,可以实现辅助打分功能。In order to improve the fairness and justice of Sanda auxiliary scoring,a method of Sanda action recognition and evaluation based on bone key points is proposed.Firstly,ten kinds of Sanda move-ments are collected to construct experimental data sets,and the key frames of the video are obtained,and the images are denoised.Secondly,the coordinates of key points of human skeleton are extracted by HRNet-DSC-CBAM human posture estimation method based on YOLOv5s-CBAM target detection.Finally,ST-GCN motion recognition method is used to identify the motion,and dynamic time warping(DTW)algorithm is used to complete the motion evaluation.The experimental results show that the method has a good recognition effect in 10 kinds of Sanda movements in the self-made data set,and can realize the auxiliary scoring function.

关 键 词：目标检测 动作识别 姿态估计 动态时间规整 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]