Künzell Stefan, Knoblich Anna, Stippler Annika
Institute of Sports Science, University of Augsburg, Augsburg, Germany.
Front Sports Act Living. 2025 Jan 17;6:1482701. doi: 10.3389/fspor.2024.1482701. eCollection 2024.
Extended feedback on knowledge of performance in sports techniques is very challenging and requires a high level of expertise. This poses a significant problem for experiments on providing extended feedback, as it is essential to ensure that the "correct" feedback is given for it to be effective. In this study, we investigate whether the correct feedback can be determined based on kinematic data. Ten participants and one model were recorded during shot put using a Motion Capturing (MoCap) system and simultaneously captured on video. The videos were analysed by two experts, and the two most critical errors were noted. By qualitatively comparing the deviations of the participants from the model, the experts' error feedback was identified in the motion curves of the MoCap system. The expert feedback for two participants was sealed in an envelope. In a qualitative analysis of the motion data, the error feedback was then determined and subsequently compared with the experts' feedback. These error feedbacks largely matched. It was shown that, in principle, it is possible to extract errors from the kinematic angle and distance curves of the movement. This study opens the door to an automated version of the qualitative assessment of movements by AI. Further research can now focus on the topic of conveying AI-generated feedback. This could then also provide a valid foundation for experiments on the effects of knowledge of performance.
提供关于运动技术表现知识的详细反馈极具挑战性，需要高水平的专业知识。这给提供详细反馈的实验带来了重大问题，因为要确保反馈有效，给出“正确”的反馈至关重要。在本研究中，我们调查是否可以基于运动学数据确定正确的反馈。在铅球投掷过程中，使用动作捕捉（MoCap）系统记录了10名参与者和1名模型，并同时进行视频拍摄。由两名专家对视频进行分析，并记录了两个最关键的错误。通过定性比较参与者与模型的偏差，在MoCap系统的运动曲线中识别出专家的错误反馈。将两名参与者的专家反馈密封在一个信封中。在对运动数据的定性分析中，随后确定错误反馈，并将其与专家反馈进行比较。这些错误反馈基本匹配。结果表明，原则上可以从运动的运动学角度和距离曲线中提取错误。本研究为人工智能对运动进行定性评估的自动化版本打开了大门。现在，进一步的研究可以集中在传达人工智能生成的反馈这一主题上。这也可以为关于表现知识效果的实验提供有效的基础。
