IMU-based judo ukemi detection approach for quantitative assessment
Journal
IET Conference Proceedings
Journal Volume
2025
Journal Issue
15
Start Page
412
End Page
414
ISSN
2732-4494
Date Issued
2025-08
Author(s)
Abstract
The ukemi technique is a critical protective measure that ensures athletes can fall safely in judo and is a fundamental skill that every beginner must master. Ukemi not only effectively protects athletes’ heads and trunks but also helps prevent severe injuries resulting from direct impact with the tatami. However, traditional methods for evaluating movement performance primarily rely on coaches' visual judgment and video analysis, which are inherently subjective and susceptible to errors due to visual obstructions. To address these limitations, this study developed an objective and automated classification system for the quantitative assessment of judo ukemi. Data collection was conducted using six inertial measurement units (IMUs) mounted on the athlete's head, arms, waist, and thighs. Motion-related features were extracted to facilitate the model's automatic detection of ukemi across different throwing techniques. A total of 27 participants (22 males and 5 females) were recruited for the experiments, where ukemi was performed in response to five different judo techniques, with each technique executed three times, resulting in a total of 15 ukemi executions. This study employed a CNN-LSTM model as the primary training model, combined with a fixed overlapping sliding window and a majority voting smoothing strategy to optimize the prediction results. Experimental results demonstrate that the proposed system achieves a precision of over 90% in detecting the ukemi technique, with the highest precision of 92.74% obtained when the window length is set to 128 samples. These findings provide a reliable basis for the objective evaluation of judo ukemi and validate the feasibility of using IMU sensors for motion analysis. © The Institution of Engineering & Technology 2025.
Event(s)
2025 International Conference on Applied System Innovation, ICASI
Subjects
JUDO
MACHINE LEARNING
UKEMI
WEARABLE INERTIAL MEASUREMENT UNITS
Publisher
Institution of Engineering and Technology (IET)
Description
Conference city:Tokyo
Conference date:22 April 2025 - 25 April 2025
Conference code:211992
Conference date:22 April 2025 - 25 April 2025
Conference code:211992
Type
conference paper