Development of a throwing skill evaluation system using machine learning

2023 Fiscal Year Final Research Report

• Project/Area Number: 19K24289
• Research Category: Grant-in-Aid for Research Activity Start-up
• Allocation Type: Multi-year Fund
• Review Section: 0909:Sports sciences, physical education, health sciences, and related fields
• Research Institution: Tohoku Gakuin University
• Principal Investigator: YOSIHDA Yudai 東北学院大学, 人間科学部, 准教授 (20754683)
• Project Period (FY): 2019-08-30 – 2024-03-31
• Keywords: 機械学習 / スポーツ技能評価 / 投動作

Outline of Final Research Achievements

The aim of this study was to develop a method for assessing throwing ability using machine learning.
University students who were members of a baseball team (considered to have high throwing ability) and university students who had no baseball experience (considered to have low throwing ability) performed a target-hitting task with a softball. The throwing motion was recorded by a digital video camera from a direction perpendicular to the throwing direction. From these videos, the still images were extracted at the moment when the non-throwing side leg made contact with the ground, which is a characteristic phase of the throwing motion. These still images were divided into training and test datasets, and a classification test using machine learning (ResNet-152) was performed. The results showed that the model achieved a high classification accuracy. That suggests the potential of using machine learning to discriminate between different levels of throwing skill.

Free Research Field

スポーツ科学

Academic Significance and Societal Importance of the Research Achievements

スポーツ技能を機械学習により評価する研究では，何かしらのセンサーやデバイスを用いて実施している研究が多い．一方，本研究では画像のみでスポーツ技能の評価を行っている．画像のみを用いてほぼ機械学習のみでスポーツ技能評価ができる可能性を示したことは，スポーツ科学として意義があると考えられる．また，体育・スポーツの現場の運用を考慮すると，センサーを用いる場合は参加者数分のセンサーを用意する必要があり運用面の手間も大きい．近年のスマートフォンやカメラの技術的な進歩によって，試技の映像を撮ることはとても簡便になっている．これらの観点から，現場活用を視野に入れている本研究は社会的に意義があると考えられる．