Sport Action Evaluation Based on Human Pose Estimation: A Case of Evaluating Golf Swing Action

Sport action evaluation is a crucial segment in sports teaching. With the advancement of human pose estimation, the extraction of human body coordinates from images has become increasingly feasible. This paper proposes a sport action evaluation approach based on human pose estimation for comprehensive quantitative action evaluation. Firstly, key postures and features are designated by professional coaches from the standard action video according to sport action characteristics. Then, key postures and features are extracted based on human pose estimation technology to obtain actual features values from the test action video. Finally, a comprehensive action evaluation approach is be constructed with the use of Fuzzy Comprehensive Evaluation (FCE), which outputs an overall score to achieve quantitative evaluation. To verify the effectiveness of the approach, this paper established a dataset comprising 100 golf swing videos labeled with the hitting distances and conducted quantitative experiments on the dataset. The results show a positive correlation between the scores of the golf swing action and the actual hitting distance, which has the practical significance of the evaluation outcomes. The proposed approach in this paper could be applied to sport action evaluation which emphasize action standardization, offering a novel perspective for sport teaching.