Full-Body Golf Swing Kinematic Reconstruction From a Smartwatch IMU

This study addresses the challenge of capturing full-body golf swing kinematics without the need for cumbersome laboratory-grade equipment. The authors introduce the Wrist-IMU Temporal Kinematic Network (WIT-KinNet), a deep learning model designed to infer full-body joint rotations from the inertial measurement unit (IMU) data of a single wrist-worn smartwatch. The framework utilizes modality-specific embeddings for acceleration, angular velocity, and orientation, combined with temporal kinematic encoder blocks to capture both long-range swing-phase dependencies and local motion dynamics.

The model was trained and validated using a dataset of 36 golfers of varying skill levels, covering seven different club types and three swing amplitudes. By comparing the WIT-KinNet predictions against ground-truth data from an optical motion capture system, the researchers demonstrated that a single wrist sensor is sufficient to reconstruct complex full-body movements. Key metrics, including pelvic and upper torso rotation, as well as the X-factor and S-factor, showed strong temporal correlations (r ≥ 0.96) with the gold-standard measurements.

Traditional golf swing analysis is often restricted to laboratory settings due to the reliance on multi-camera optical systems or multi-sensor wearable arrays that are difficult to calibrate and use during actual play. By enabling full-body kinematic reconstruction from a single, commercially available smartwatch, this research provides a practical, non-invasive tool for golfers and coaches. This advancement facilitates objective, data-driven feedback on swing technique in real-world environments, potentially democratizing access to professional-grade biomechanical analysis.