Category
Technology/equipment
Document Type
Paper
Abstract
Accurate ground reaction force (GRF) measurement is essential for in-depth golf swing analysis, but force plates are costly and require complex installation. This study develops an AI-based GRF estimation method using a Bi-LSTM model. Motion capture and force plate data from 1,032 swings by 344 golfers were used to train the model, while data from 29 unseen golfers were reserved for validation to prevent overfitting. A total of 864 models were trained using cross-validation and grid search, with the best-performing model selected based on ICC. The model achieved high accuracy (ICC up to 0.983), particularly for lead foot vertical GRF. This approach provides a scalable, cost-effective solution for movement analysis, with potential applications beyond golf.
Recommended Citation
Mori, Kanji
(2025)
"Estimation of the Ground Reaction Forces During Golf Swing Using Recurrent Neural Networks,"
ISBS Proceedings Archive: Vol. 43:
Iss.
1, Article 60.
Available at:
https://commons.nmu.edu/isbs/vol43/iss1/60
