This study examined whether a low-cost, two-dimensional (2D) video camera combined with supervised machine learning (ML) models could accurately predict the Dynamic Strength Index (DSI), a key indicator of neuromuscular performance. A total of 263 healthy participants performed countermovement jumps (CMJ) and isometric mid-thigh pulls (IMTP). Unlike the conventional method where both ballistic and isometric peak forces are measured using a force plate, in this study the isometric force was assessed using a back and leg dynamometer, while ballistic force during CMJ was estimated from 2D video data. Six spatiotemporal features extracted from the video, together with height and weight, were used to train multiple supervised regression models. Ground truth validation of the video-based ballistic force estimation was performed using force plate data. Gaussian Process Regression and Neural Networks achieved the highest prediction accuracy (R² > 0.92, RMSE < 0.06), demonstrating excellent agreement with laboratory-derived values. These findings highlight the potential of combining simple video systems with ML algorithms to estimate DSI accurately and affordably. The proposed method provides a scalable and non-invasive alternative to lab-grade equipment, enabling broader access to neuromuscular diagnostics and performance monitoring, particularly in youth, community sports, and field-based environments.
Keywords: 2D video analysis; Dynamic strength index (DSI); Force plate; Ground reaction force (GRF); Machine learning; Sports performance.
© 2025. The Author(s).