2022 Fiscal Year Final Research Report
Criterion-referenced measurement and evaluation of movement skills using deep machine learning
| Project/Area Number |
21K19697
|
|---|
| Research Category |
Grant-in-Aid for Challenging Research (Exploratory)
|
| Allocation Type | Multi-year Fund |
|---|
| Review Section |
Medium-sized Section 59:Sports sciences, physical education, health sciences, and related fields
|
|---|
| Research Institution | University of Tsukuba |
|---|
Principal Investigator |
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
安藤 梢 筑波大学, 体育系, 助教 (30904007)
|
|---|
| Project Period (FY) |
2021-07-09 – 2023-03-31
|
| Keywords | 運動軸推定 / 運動姿勢推定 / 深層学習 / 動作技能 / 運動能力 / 達成度評価 / 項目反応理論 |
|---|
| Outline of Final Research Achievements |
We attempted to construct a new method for measuring and evaluating motor skills using deep learning. We attempted to solve a step-by-step research problem by deep learning physical movement features to measure motor skills as movement skills from movement videos and to elucidate an algorithm for evaluating the level of performance. (1) The movement techniques and performance criteria perceived by skilled athletes were identified. (2) Items and evaluation criteria for assessing the performance level of movement techniques were clarified. (3) Applying deep learning techniques, we revealed a body axis estimation algorithm that displays the body axis from the measurement points of the limb trunks output from the posture estimation algorithm.
|
| Free Research Field |
体育測定評価学，スポーツ統計学
|
| Academic Significance and Societal Importance of the Research Achievements |
【学術的意義】スポーツ映像への深層学習アルゴリズム構築技術の適用は，早急な解決が必要不可欠な研究課題である．(1) 運動映像から運動能力を測定する深層学習アルゴリズムの構築，(2) 動作技能項目の達成度評価基準を分析することで，深層学習のためのKPIの解明，(3) スポーツの専門家が有する暗黙知である「匠の技」の見える化．【社会的意義】データテクノロジーと深層学習手法の適用により，スポーツ映像から動作技能を達成度評価するアルゴリズムの構築は，スポーツや体育でのICT教育の発展基盤となる．
|
