Bridging the lab-to-field gap using machine learning: a narrative review

Area of Science:

• Sports Biomechanics
• Machine Learning Applications
• Human Movement Analysis

Background:

• Bridging the lab-to-field gap in sports biomechanics is crucial for practical applications.
• Current machine learning (ML) applications are hindered by the lack of large, high-quality datasets from on-field analysis.
• Existing kinematic and kinetic data primarily originate from laboratory-based motion capture, not readily available wearable or video-based systems.

Purpose of the Study:

• To summarize recent advancements in ML applications for sports biomechanics, focusing on overcoming the lab-to-field gap.
• To explore methods for repurposing existing motion capture data for ML-driven on-field motion analysis.
• To derive guidelines for effective ML implementation in biomechanics, addressing dataset size, algorithm selection, and data variability.

Main Methods:

• Review of recent advancements in machine learning applications within sports biomechanics.
• Summarization of methods to re-purpose laboratory motion capture data for ML on-field analysis.
• Overview of current ML applications to identify best practices and derive guidelines.

Main Results:

• Identified challenges in ML for sports biomechanics include data scarcity and lack of standardized guidelines.
• Proposed methods to adapt existing motion capture data for ML-based on-field analysis.
• Highlighted the need for large-scale, high-quality databases for both traditional and emerging sensor technologies.

Conclusions:

• Repurposing existing data and developing clear guidelines are essential for advancing ML in sports biomechanics.
• Progress in ML applications requires addressing data limitations and establishing best practices for algorithm selection and dataset characteristics.
• This work aims to facilitate research towards effectively bridging the lab-to-field gap in sports performance analysis.