Expert Comment Generation Considering Sports Skill Level Using a Large Multimodal Model with Video and...

Area of Science:

• Sports Science
• Artificial Intelligence
• Biomechanics

Background:

• Personalized skill assessment and feedback are vital for athletic improvement.
• Existing methods often overlook spatial-temporal movement data and skill-level tailoring in feedback generation.

Purpose of the Study:

• To develop a novel approach for generating skill-level-aware expert comments for athletes.
• To enhance the personalization and actionability of feedback in sports training.

Main Methods:

• Utilized a Spatial-Temporal Attention Graph Convolutional Network (STA-GCN) to extract motion features and classify skill levels.
• Integrated skill level classification and motion features into a Large Multimodal Model (LMM) for comment generation.
• Combined video analysis with spatial-temporal motion dynamics for comprehensive feedback.

Main Results:

• The proposed method successfully generates detailed, context-specific expert comments.
• Feedback is tailored to the learner's specific skill level, addressing limitations of prior research.
• The integration of motion features enhances the LMM's ability to provide actionable insights.

Conclusions:

• The novel approach effectively generates expert comments, offering valuable guidance for athletes.
• This method represents a significant advancement in personalized sports training and skill acquisition.
• The skill-level-aware feedback system has broad applicability across various sports and skill levels.