数据与计算发展前沿 ›› 2026, Vol. 8 ›› Issue (3): 181-190.

doi: 10.11871/jfdc.issn.2096-742X.2026.03.015

• 技术与应用 • 上一篇    下一篇

基于多流图-时序融合网络的警察训练动作智能评分方法

张培晶1(),严佳欣2,王晓璇3,李俊杰2,*(),曾云飞1   

  1. 1 中国人民公安大学， 信息网络安全学院， 北京 100038
    2 兴质（北京）科技研究院有限公司， 北京 102629
    3 中国人民公安大学， 警体战训学院， 北京 100038
  • 收稿日期:2025-10-31 出版日期:2026-06-20 发布日期:2026-06-18
  • 通讯作者: 李俊杰
  • 作者简介:张培晶，中国人民公安大学，副教授，硕士，主要研究方向为计算机视觉、人工智能等。
    本文中负责概念构思，方法设计，论文审阅与编辑。
    ZHANG Peijing is an associate professor at People's Public Security University of China and holds a master’s degree. His current research interests include computer vision, artificial intelligence.
    In this paper, he is mainly responsible for conceptualization, methodology design, and review and editing the manuscript.
    E-mail: zhangpeijing@ppsuc.edu.cn|李俊杰，兴质（北京）科技研究院有限公司，硕士，主要研究方向为信号处理与人工智能算法在可穿戴智能设备中的应用。
    本文中负责数据处理与算法实现。
    LI Junjie holds a master's degree and works at Xing-Zhi(Beijing) Technology Research Institute Co., Ltd. His research interests include signal processing and artificial intelligence algorithms for wearable smart devices.
    In this paper, he is responsible for data processing and algorithm implementation.
    E-mail: ljj12393@163.com

Intelligent Scoring Method for Police Training Action Based on Multi-Stream Graph-Temporal Fusion Network

ZHANG Peijing1(),YAN Jiaxin2,WANG Xiaoxuan3,Li Junjie2,*(),ZENG Yunfei1   

  1. 1 College of Informatics and Cyber Security, People’s Public Security University of China, Beijing 100038, China
    2 Xing-zhi (Beijing) Technology Research Institute Co., Ltd., Beijing 102629, China
    3 College of Police Law Enforcement Abilities Training, People’s Public Security University of China, Beijing 100038, China
  • Received:2025-10-31 Online:2026-06-20 Published:2026-06-18
  • Contact: Li Junjie

RichHTML

0

PDF

11

摘要：

【目的】针对警察实战训练考核依赖人工评分所存在的动作质量评价主观性强、效率低、量化打分困难等问题，提出一种基于多流图-时序融合网络（MS-GTFN）的动作智能评分方法，为警察自主开展实战训练提供质量评价参考。 【方法】首先，构建四类动作特征流——关节流、骨骼流、关节运动流与骨骼运动流，综合编码姿态结构与动态演化信息；其次，并行使用图卷积网络（GCN）与时间卷积网络（TCN）提取动作的时空融合特征；随后，引入通道注意力（CA）和空间自注意力（SSA）模块，进一步增强对关键特征的利用能力；最后，通过输出层多层感知机（MLP）实现动作评分预测。 【结果】方法在自建警察训练数据集上进行训练与验证，最终在评分拟合任务中达到良好效果，MAE、MSE与决定系数R2分别为0.4553、0.3507与0.9306。 【结论】方法在训练动作特征融合与评分准确性方面具有显著优势,可以为警察训练动作质量智能评分提供较为准确的结果。

关键词: 图卷积网络, 时间卷积网络, 多流融合, 注意力模块, 警察实战训练, 动作质量评分

Abstract:

[Purpose] An intelligent scoring method training action based on a Multi-Stream Graph-Temporal Fusion Network (MS-GTFN) is proposed to address the issues of subjective evaluation, low efficiency, insufficient standardization, and difficulty in quantifying action quality in police training assessments that rely on manual scoring. This method provides a reliable quality evaluation reference for police officers to conduct independent training. [Methods] First, four types of spatiotemporal feature streams—joint stream, bone stream, and their corresponding motion streams—are constructed to comprehensively encode both structural and dynamic characteristics of movements. Second, Graph Convolutional Networks (GCNs) and Temporal Convolutional Networks (TCNs) are employed in parallel to extract spatiotemporal fusion features of actions. Subsequently, channel attention (CA) and spatial self-attention (SSA) modules are introduced to further enhance the model’s capability to focus on key features. Finally, a multi-layer perceptron (MLP) is used to predict action score. [Results] The proposed model is trained and validated on our self-built police training dataset. Experimental results demonstrate promising performance in motion action score prediction tasks, achieving an MAE of 0.4553, an MSE of 0.3507, and an R2 of 0.9306. [Conclusions] The method offers significant advantages in the fusion of training action features and scoring accuracy, providing more reliable support for intelligent scoring of police training action quality.

Key words: graph convolutional networks, temporal convolutional networks, multi-stream fusion, attention module, police force training, scoring of action quality