Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Videos

Sports action estimation and recognition based on dynamic spatiotemporal graph convolution.

Fengling Zhang1, Huiyang Xiao2, Weitao Guo3

  • 1School of Sports Engineering Innovation and Interdisciplinary Studies, Guangdong University of Technology, GuangZhou, 510006, China.

BMC Sports Science, Medicine & Rehabilitation
|July 3, 2026
PubMed
Summary
This summary is machine-generated.

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Should a Dorsal or Anal Scale Be Used for Microcharacteristics Analysis in <i>Coreius guichenoti</i>?

Animals : an open access journal from MDPI·2026
Same author

Clinical Spectrum, independent risk factors, and treatment outcomes of pediatric arrhythmias: a multicenter retrospective analysis in Xinjiang.

Frontiers in pediatrics·2026
Same author

Piezoelectric-Based Vibration Energy-Harvesting for Bladed Disks: Modeling and Comparative Performance Analysis of Interface Circuits.

Sensors (Basel, Switzerland)·2026
Same author

Diagnostic performance of wearable acoustic-electrocardiographic indices for screening secundum atrial septal defect in children.

Frontiers in pediatrics·2026
Same author

Barriers to and facilitation of long-term physical exercise participation among Chinese adolescents with autism spectrum disorder: a parental perspective.

Frontiers in psychology·2026
Same author

Effects of a dual-team collaboration model guided by chain management on ECMO initiation and clinical outcomes in critically ill patients: a quasi-experimental study.

Frontiers in medicine·2026
Same journal

Comparative effects of pilates-based interventions on functional mobility, balance, fatigue, and quality of life in people with multiple sclerosis: a systematic review and network meta-analysis.

BMC sports science, medicine & rehabilitation·2026
Same journal

Associations among age, playing position, dribbling, agility, and passing skills in elite youth football players graduating from a high-performance academy: an observational study.

BMC sports science, medicine & rehabilitation·2026
Same journal

Self-talk levels in male individuals aged 18 and over who actively play soccer.

BMC sports science, medicine & rehabilitation·2026
Same journal

Postural stability under sensory challenge in Masters weightlifters and its association with weightlifting performance and peak power: a cross-sectional observational study.

BMC sports science, medicine & rehabilitation·2026
Same journal

Acute effects of cryotherapy duration on upper extremity muscle strength, flexibility, endurance, and balance: a randomized trial.

BMC sports science, medicine & rehabilitation·2026
Same journal

Blunted exercise-induced hypoalgesia in people with knee pain after a brief low-intensity warm-up: a split-body study with and without blood flow restriction.

BMC sports science, medicine & rehabilitation·2026
See all related articles

This study introduces advanced graph convolutional network models for precise sports action pose estimation and motion recognition. These models significantly improve accuracy and efficiency in analyzing athlete movements for enhanced sports performance and research.

Area of Science:

  • Computer Science
  • Sports Science
  • Artificial Intelligence

Background:

  • Accurate estimation and recognition of sports movements are crucial for performance analysis and research.
  • Existing methods may lack the precision and efficiency required for complex athletic actions.

Purpose of the Study:

  • To develop and evaluate novel models for accurate sports action pose estimation and motion recognition.
  • To enhance the understanding and analysis of athletic movements through improved computational models.

Main Methods:

  • An improved graph convolutional network (GCN) was utilized for action pose estimation.
  • A dynamic spatiotemporal graph convolutional network (DST-GCN) was designed for motion recognition.
  • Performance was evaluated based on joint detection rates, pose point error, recognition accuracy, and computational efficiency.
Keywords:
Convolutional networkGraph convolutional networkMotion recognitionPose estimationSports videos

Related Experiment Videos

Main Results:

  • The action pose estimation model achieved a joint detection percentage of 0.986 and a success rate of 0.975, with a maximum average pose point position error of 0.25.
  • The DST-GCN model demonstrated superior recognition accuracy (0.959 average, 0.96 Top-1, 0.97 Top-5) and efficiency (192.988 FPS, 98.754 FLOPs).
  • The models showed robustness in identifying behavior across different perspectives and individuals.

Conclusions:

  • The developed GCN-based models provide highly accurate and efficient solutions for sports movement analysis.
  • These advancements contribute to improved sports performance insights and the broader field of sports research.
  • The study highlights the potential of deep learning, specifically GCNs, in understanding complex human actions in sports.