Automated Assessment of Developmental Stage of Overhand Throwing Skill

The trade-off between speed and accuracy in scoring process-oriented tests for fundamental movement skills (FMS) has always been challenging for a screening project. The aim of this study was the feasibility of using wearable inertial measurement units (IMU) and artificial intelligence algorithms to automatically assessment of FMS. ۱۲۳ overhand throwings were performed by children aged ۴ to ۱۰ years (age = ۷±۱.۸۴) (۵۳% = boys). Three IMU (Shokofa Tavan Vira) sent signals of angular velocity, linear acceleration of the preferred hand, non-predominant ankle, and lumbar region of the children. Each performance was scored according to the criteria of the third edition test of The Gross Motor Development (TGMD-۳) by reviewing the video of the performed skills. The "k nearest neighbor" algorithm was used for automatic data classification. The minimum difference between test signals and training signals was calculated and classified. Two issues were assessed: false acceptance, in which an “incorrect” performance was classified as “correct”; and false rejection where a “correct” performance was classified as “incorrect”. The classification accuracy of the K-nearest neighbor (KNN) algorithm was ۸۵%. The automatic scoring algorithm also correctly classified ۹۳%, ۷۸%, ۹۳%, and ۷۶% in criteria ۱ to ۴, respectively. Low-back IMU data analysis shows the model's accuracy of ۷۵%. Further, the total scoring time was reduced from ۵ minutes to less than ۳۰ seconds. The use of artificial intelligence in the signal processing of only three IMU was a reliable and practical method for the assessment of FMS. This approach means the monitoring and evaluation of children's movement skills can be objective. In addition, while maintaining relative accuracy, the time involved in the process-oriented analysis of FMS for research, clinical, sports, and educational purposes was reduced entirely.

Keywords:

Test of Gross Motor Development , Wearable Inertial Measurement Unit , Artificial Intelligence , motor development , Automatic Assessment