Counterfactual Explanation-Based Badminton Motion Guidance Generation Using Wearable Sensors

TL;DR

This work tackles how to generate personalized badminton motion guidance using wearable sensors and counterfactual explanations to bridge the gap between novice and expert performance. It introduces a LatentCF framework comprising a transformer-based stroke-quality classifier, a counterfactual explainer operating in latent space, and a Unity-based visualization to produce expert-like, non-imitation guidance from a novice's motion, trained and validated on the MultiSenseBadminton dataset with $L_1$, $L_2$, $L_ ext{inf}$, DTW, and other motion-specific metrics. Results show that CF-generated motions preserve the essence of the original movements while improving stroke quality and achieving closer alignment to expert motions than direct imitation, with high validity and plausible motion characteristics as measured by LOF, IF, and OCSVM indicators. The approach holds promise for scalable, personalized sports training and could extend to other domains such as robotics and skill transfer, where counterfactual-grounded guidance can augment coaching without expert supervision.

Abstract

This study proposes a framework for enhancing the stroke quality of badminton players by generating personalized motion guides, utilizing a multimodal wearable dataset. These guides are based on counterfactual algorithms and aim to reduce the performance gap between novice and expert players. Our approach provides joint-level guidance through visualizable data to assist players in improving their movements without requiring expert knowledge. The method was evaluated against a traditional algorithm using metrics to assess validity, proximity, and plausibility, including arithmetic measures and motion-specific evaluation metrics. Our evaluation demonstrates that the proposed framework can generate motions that maintain the essence of original movements while enhancing stroke quality, providing closer guidance than direct expert motion replication. The results highlight the potential of our approach for creating personalized sports motion guides by generating counterfactual motion guidance for arbitrary input motion samples of badminton strokes.

Figures (1)

• Figure 1: Motion guidance generation framework