混合注意力机制下竞技体育运动违规动作识别算法

Recognition Algorithm for Violation Actions in Competitive Sports under Hybrid Attention Mechanism

摘要

由于违规动作多为瞬时行为, 易被大量合规动作的冗余信息掩盖, 不同距离下的特征区分度不佳, 导致识别准确率不高。 为此, 引入混合注意力机制, 对竞技体育运动违规动作识别算法进行研究。 首先利用背景减除法定位运动员区域, 结合高斯滤波与下采样技术构建尺度空间, 以适配不同距离下的动作特征。 然后采用Hessian 行列式提取时空域与多尺度特征并通过双向传输融合高低层特征, 以提高特征的区分度。 空间注意力聚焦于肢体关键部位, 时序注意力锁定违规动作的关键帧, 通道注意力强化有效特征通道, 三者协同作用以输出增强特征。 最后通过端到端的训练优化模型, 结合序列距离判断区分相似动作, 实现违规动作的识别。实验结果表明, 该算法能有效分离相似动作特征, 违规动作识别准确率超过 99% , 平均识别准确率达99. 81% ,均优于对比算法, 具有较大的应用价值。

Abstract

Due to the fact that most of the violations are instantaneous actions, they are easily obscured by redundant information from a large number of compliant actions, resulting in poor feature discrimination at different distances and low recognition accuracy. Therefore, a mixed attention mechanism is proposed to conduct research on the recognition algorithm of illegal movements in competitive sports. Background subtraction is used to locate the athlete’s area, and Gaussian filtering and downsampling techniques are combined to construct a scale space to adapt to motion features at different distances. Hessian determinant is used to extract spatiotemporal and multi-scale features, and high-level features are fused through bidirectional transmission to improve feature discrimination. Spatial attention focuses on key parts of the limbs, temporal attention locks in keyframes of illegal actions, channel attention enhances effective feature channels, and the three work together to output enhanced features. Finally, the end-to-end training optimization model is used, combined with sequence distance judgment to distinguish similar actions and achieve the recognition of illegal actions. The experimental results show that the algorithm can effectively separate similar action features, with a recognition accuracy of over 99% for illegal actions, an average recognition accuracy of 99. 81% , which are better than the comparative algorithms and have greater application value.

关键词

混合注意力机制 / 竞技体育 / 违规动作 / 动作识别 / 空间通道注意力

Key words

hybrid attention mechanism / competitive sports / illegal actions / action recognition / spatial and channel attention

参考文献

[1] 古学茹, 张备伟, 丁雯. 基于多流自适应时空图卷积网络的人体行为识别[J]. 南京师大学报(自然科学版), 2025, 48(3): 112-119.

[2] GU X R, ZHANG B W, DING W. Human Behavior Recognition Based on Multi-Stream Adaptive Spatio-Temporal Graph Convolutional Networks[J]. Journal of Nanjing Normal University (Natural Science Edition), 2025, 48(3): 112-119.

[3] 陶峰, 李燕苹, 王瑞. 基于图卷积网络改进的人体动作识别模型[J]. 电子测量技术, 2023, 46(8): 59-64.

[4] TAO F, LI Y P, WANG R. Improved Human Action Recognition Model Based on Graph Convolutional Networks[J]. Electronic Measurement Technology, 2023, 46(8): 59-64.

[5] 胡明, 许佳炜, 赵立军, 等. 基于小样本和随机化的跨域人体动作泛化识别模型[J]. 计算机应用研究, 2025, 42(3): 849-855.

[6] HU M, XU J W, ZHAO L J, et al. Cross Domain Human Action Generalization Recognition Model Based on Small Sample Size and Randomization[J]. Application Research of Computers, 2025, 42(3): 849-855.

[7] 赵登阁, 智敏. 基于时空注意图卷积的人体动作识别[J]. 计算机应用与软件, 2024, 41(7): 165-170,254.

[8] ZHAO D G, ZHI M. Action Recognition Based on Spatial-Temporal Attention Graph Convolution Neural Network[J]. Computer Applications and Software, 2024, 41(7): 165-170,254.

[9] 张传宗, 王冬子, 郭政鑫, 等. 基于双流融合网络的非接触式 IR-UWB 人体动作识别方法[J]. 计算机科学, 2025, 52(1): 221-231.

[10] ZHANG C Z, WANG D Z, GUO Z X, et al. Contact-Free IR-UWB Human Motion Recognition Based on Dual-Stream Fusion Network[J]. Computer Science, 2025, 52(1): 221-231.

[11] 李新春, 曾仕豪. 基于改进粒子群算法的 UWB 雷达人体动作识别研究[J]. 重庆邮电大学学报(自然科学版), 2024, 36(2): 268-276.

[12] LI X C, ZENG S H. Research on UWB Radar Human Motion Recognition Based on Improved Particle Swarm Optimization Algorithm[J]. Journal of Chongqing University of Posts and Telecommunications (Natural Science Edition), 2024, 36(2): 268-276.

[13] 李晶晶, 黄章进, 邹露. 基于运动引导图卷积网络的人体动作识别[J]. 计算机辅助设计与图形学学报, 2024, 36(7): 1077-1086.

[14] LI J J, HUANG Z J, ZOU L. Motion-Guided Graph Convolutional Network for Human Action Recognition[J]. Journal of Computer-Aided Design & Computer Graphics, 2024, 36(7): 1077-1086.

[15] 廖民玲. 基于显著性特征的多视角人体动作图像识别研究[J]. 现代电子技术, 2024, 47(24): 143-147.

[16] LIAO M L. Research on Multi-Perspective Human Action Image Recognition Based on Significant Features[J]. Modern Electronics Technique, 2024, 47(24): 143-147.

[17] 石峰. 基于动态时间规整耦合3D 运动历史图像的人体动作识别算法[J]. 传感技术学报, 2024, 37(11): 1937-1945.

[18] SHI F. Human Motion Recognition Algorithm Based on Dynamic Time Warping Coupled with 3D Motion History Image[J]. Chinese Journal of Sensors and Actuators, 2024, 37(11): 1937-1945.

[19] 李程辉, 蒋俊正, 周芳. 基于毫米波雷达的三维特征自适应融合人体动作识别算法[J]. 雷达科学与技术, 2024, 22(5): 569-577, 586.

[20] LI C H, JIANG J Z, ZHOU F. Human Motion Recognition Algorithm with Three-Dimensional Feature Adaptive Fusion Based on Millimeter-Wave Radar[J]. Radar Science and Technology, 2024, 22(5): 569-577, 586.

[21] 王杨, 许佳炜, 王傲, 等. 基于 CSI 实例标准化的域泛化人体动作识别模型[J]. 通信学报, 2024, 45(6): 196-209.

[22] WANG Y, XU J W, WANG A, et al. Domain-Generalization Human Activity Recognition Model Based on CSI Instance Normalization[J]. Journal on Communications, 2024, 45(6): 196-209.

[23] 陶志勇, 郭希俊, 任晓奎, 等. 基于多通道特征融合的人体动作识别方法[J]. 工程科学与技术, 2025, 57(1): 68-79.

[24] TAO Z Y, GUO X J, REN X K, et al. Human Action Recognition Method Based on Multi-Channel Fusion[J]. Advanced Engineering Sciences, 2025, 57(1): 68-79.

[25] 王丽芳, 吴荆双, 尹鹏亮, 等. 基于注意力机制和能量函数的动作识别算法[J]. 计算机应用, 2025, 45(1): 234-239.

[26] WANG L F, WU J S, YIN P L, et al. Action Recognition Algorithm Based on Attention Mechanism and Energy Function[J]. Journal of Computer Applications, 2025, 45(1): 234-239.

[27] 程娜娜, 张荣芬, 刘宇红, 等. 基于 ConvGRU 和注意力特征融合的人体动作识别[J]. 光电子·激光, 2023, 34(12): 1298-1306.

[28] CHENG N N, ZHANG R F, LIU Y H, et al. Human Motion Recognition Based on ConvGRU and Attention Feature Fusion[J]. Journal of Optoelectronics·Laser, 2023, 34(12): 1298-1306.

[29] 郭宗洋, 刘立东, 蒋东华, 等. 基于语义引导神经网络的人体动作识别算法[J]. 图学学报, 2024, 45(1): 26-34.

[30] GUO Z Y, LIU L D, JIANG D H, et al. Human Action Recognition Algorithm Based on Semantics Guided Neural Networks[J]. Journal of Graphics, 2024, 45(1): 26-34.

[31] 黄斌, 王寅昊. 激光传感器和机器视觉的复杂运动动作识别方法[J]. 激光杂志, 2023, 44(10): 217-224.

[32] HUANG B, WANG Y H. Complex Motion Recognition Method Based on Laser Sensor and Machine Vision[J]. Laser Journal, 2023, 44(10): 217-224.

[33] 王子威, 郭苗苗. 多视角手部肌肉疲劳动作智能识别方法仿真[J]. 计算机仿真, 2024, 41(1): 238-242.

[34] WANG Z W, GUO M M. Simulation of Intelligent Recognition Method for Hand Muscle Fatigue Actions from Multiple Perspectives[J]. Computer Simulation, 2024, 41(1): 238-242.

[35] 林远强, 郜辉, 王鹏, 等. 引入级联通道注意力的轻量化人体姿态估计[J]. 计算机工程与应用, 2024, 60(13): 219-227.

[36] LIN Y Q, GAO H, WANG P, et al. Lightweight Human Pose Estimation with Cascaded Channel Attention[J]. Computer Engineering and Applications, 2024, 60(13): 219-227.

[37] 申超凡, 熊风光, 孔煜, 等. 基于自注意力和曲率的点云生成对抗网络[J]. 计算机工程与设计, 2025, 46(7): 1890-1897.

[38] SHEN C F, XIONG F G, KONG Y, et al. Point Cloud Generation Adversarial Network Based on Self-Attention and Curvature[J]. Computer Engineering and Design, 2025, 46(7): 1890-1897.

[39] 余本功, 石中玉. 深层注意力和两阶段融合的图文情感对比学习方法[J]. 计算机工程与应用, 2025, 61(3): 223-233.

[40] YU B G, SHI Z Y. Deep Attention and Two-Stage Fusion of Image-Text Sentiment Contrastive Learning Method[J]. Computer Engineering and Applications, 2025, 61(3): 223-233.

基金资助

四川省教育厅人文社会科学重点研究基地课题基金资助项目(XXTYCY2022D11)