Document Type : Original research papers
Department of Physical Education, Gorgan University of Agricultural Sciences and Natural Resources
Department of motor behavior, Faculty of Physical Education and Sport Sciences, University of Tehran, Tehran, Iran
Department of motor behavior, Faculty of Physical Education and Sport Sciences, University of Tehran, Tehran, Iran.
Department of Interdisciplinary Technology /Network Science and Technology, Faculty of New Sciences and Technologies, University of Tehran, Tehran, Iran.
Developing fundamental movement skills (FMS) as the building blocks of complex sports skills and daily physical activity is crucial. The mechanically optimal performance can be determined by qualitative changes in the sensitive aspects of the skill. Accurate scoring of this process is time-consuming and requires minimum training and experience. Thus, this study was designed to evaluate the feasibility of using wearable inertial units (IMUs) based on artificial intelligence algorithms (AIA) for objective assessment of catching and throwing skills.Thirteen children aged 4 to 10 years (age = 7±1.84) (boys = 53%) were asked to do at least ten repetitions of two hands ball catch and underhand throw according to the Test of Gross Motor Skills Development- third edition (TGMD-3). Trials were captured with video recording and three IMUs, simultaneously. Dynamic Time Warping (DTW) and K-Nearest Neighbor (KNN) artificial intelligence algorithms automatically classified IMU signals. The intraclass correlation coefficient (ICC) was calculated between expert scores and the artificial intelligence algorithm. All tests were done at a 95% confidence interval. The classification accuracy of the KNN algorithm (k=7) for two hand catch was 73%, ICC =0.51 (CI=0.25-0.69), and for underhand throw was 70%, ICC= 0.559, (CI=0.314-0.717). The algorithm accuracy when using lower back sensor data was 72% for the tow-hand catch and 78% for the underhand throw. The scoring time was reduced from 5 minutes per skill (in an expert-oriented way) to less than 30 seconds (using artificial intelligence). A close examination of the artificial intelligence classification revealed several aspects of performance that did not play an influential role in trials but were artificially consistent with the TGMD-3. Locating the sensor in the waist area for these two skills will save the cost and time in screening plans. This instrument assessment provides instant feedback, is portable, economical and easy-to-use, and is suitable for educational setting. In the future, more research should be conducted on IMUs' real-world applications by teachers, researchers, clinicians, and coaches.