Development and Validation of a Digital Evaluation System for a Repetitive Side-to-Side Jump Test Based on Skeletal Recognition

In Japan, the New Physical Fitness Test has been implemented annually in schools since 1998 to assess children’s physical and motor abilities. Despite its scale, measurements and data entry are often conducted manually, placing a burden on educators and limiting data reliability and utilization. This study addresses these issues by developing a low-cost MoveNet-based ICT system (hereafter referred to as ICT system) for the side-to-side jump test, which assesses agility. Fifty university students (33 males, 17 females; mean ± SD age = 19.32 ± 0.76 years) performed 243 trials. Four measurement methods were compared: the ICT system, infrared sensors, manual counting by an experimental operator, and GoPro-based reference counts obtained by frame-by-frame video review. Among 233 valid trials, Pearson correlation coefficients with the reference were r = .299 (p < .001) for the ICT system, r = .671 (p < .001) for the sensor, and r = .854 (p < .001) for manual counting. Relative to the reference, the ICT system showed a mean error (bias) of −11.55 counts (SD = 10.22), with 95 % limits of agreement ranging from −31.57 to + 8.48 counts. Bland–Altman analysis therefore revealed a consistent under-counting bias in the ICT measurements. Although the system demonstrates practical potential, further work is needed to improve the detection of rapid lateral movements. In addition, fully automatic score logging and CSV data export should be implemented to facilitate integration with school information systems.