MultiSenseBadminton: Wearable Sensor-Based Biomechanical Dataset for Evaluation of Badminton Performance

Abstract

The sports industry is witnessing an increasing trend of utilizing multiple synchronized sensors to collect player data, enabling the creation of personalized training systems with real-time feedback from multiple perspectives. Badminton could benefit from these various sensors, but there is a notable lack of comprehensive badminton action datasets for analysis and training feedback. To address this gap, this paper introduces a multi-sensor-based badminton action dataset for forehand clear and backhand drive strokes. This includes 7,763 badminton swing data from 25 players. It provides eye tracking, body tracking, muscle signals, foot pressure, detailed annotation data on stroke type, skill level, hitting sound, ball landing, hitting location, survey data, and interview data. The dataset was designed based on interviews with badminton coaches to ensure usability. The dataset includes a range of skills consisting of 12 novices, 8 intermediates, and 5 experts, providing resources for understanding biomechanics across skill levels. We validate the potential usefulness of our dataset by applying a proof-of-concept machine learning model to classify stroke type and level of expertise.