Machine Learning Classification of Poomsae Side Kick Performance
Researchers from Korea National Sport University developed machine learning (ML) models capable of classifying the execution of the Poomsae side kick, using kinematic data—joint angles, velocities, and timing—alongside physical attributes such as height, weight, and training experience. The study was published in the Journal of Sports Sciences (Taylor & Francis, 2025).
A new study applied kinematic and physical parameters to evaluate Taekwondo technique through artificial intelligence.
Researchers from Korea National Sport University developed machine learning (ML) models capable of classifying the execution of the Poomsae side kick, using kinematic data—joint angles, velocities, and timing—alongside physical attributes such as height, weight, and training experience. The study was published in the Journal of Sports Sciences (Taylor & Francis, 2025).
AI in motion analysis
The team recorded the temporal and angular patterns of the side kick’s key phases—lifting, striking, and retraction—and trained ML models to identify and classify them automatically.
Their system achieved recognition rates above 75%, demonstrating that artificial intelligence can reliably assess technical performance in Taekwondo with objective precision.
Main outcomes
The analysis found that the side kick consisted of approximately 38% of total time in the lifting-rotation phase, 16% during impact, and 46% in retraction.
Integrating athletes’ physical profiles significantly improved classification accuracy, confirming that individual body characteristics influence movement patterns and technical output.
Implications for training and judging
The research highlights the potential of ML-based tools to become objective evaluation systems for coaches, referees, and athletes.
Such technology could enhance feedback quality, provide standardized scoring frameworks, and optimize the learning process through data-driven insights.
Future studies will expand sample size, include inertial measurement sensors (IMUs), and explore deep neural networks to further refine classification accuracy.
Authors: Jong-Hoon Park, Ji-Hoon Kim, Kyu-Seok Lee, and Kyung-Hoon Kim
Institution: Department of Physical Education, Korea National Sport University (Seoul, South Korea)
Source: Journal of Sports Sciences, Taylor & Francis Online, 2025
MASTKD: Media About Sports Taekwondo. Worldwide Leader on Taekwondo Information.
About The Author
Descubre más desde MASTKD
Suscríbete y recibe las últimas entradas en tu correo electrónico.
