Understanding and Designing a Gaze-Based Interface for Table Tennis

Abstract

In fast precision sports such as table tennis, how players allocate their visual attention is an important indicator of performance. However, gaze data is difficult to interpret in real-world training settings, where eye, head, body, and scene motion make gaze signals noisy and highly context-dependent, and many existing methods are designed for controlled, offline analysis. Rather than expecting perfect predictions, we examine how gaze and multimodal sensing can support explanation, correction, personalisation, and improvement over time through a coach-facing system. To ground this design, we conducted semi-structured interviews with a coach and three table tennis players to understand their training routines and needs. The findings show that feedback should be tied to critical moments in play, combined with additional signals, translated into actionable guidance, and supported by clear explanations. Based on these insights, we propose EyeCoach, an Intelligent User Interface concept for personalised multimodal training feedback.