Samenvatting

Kicking in Australian Football (AF) has been reported as the most important skill of the game. Nevertheless, no study has investigated kicking with IMU units. These units could be used to gain kicking data through an easy and assessable way. The purpose of this study was to analyse a kick executed by a mechanical limb to evaluate the accuracy of machine learning identification of an Australian Football kick out of IMU data and evaluate the validity of a method to calculate IMU foot velocity against the reference foot velocity derived from high speed video. A mechanical limb, designed for producing drop punt kicks, performed twelve different kicks at different foot velocities. The IMU datasets were analysed with Matlab to calculate the foot velocity and create a classification system to identify kicks. The classification of correctly identified kicks provided an accuracy of 99.3% for the best produced classification system. The method to calculate IMU foot velocity demonstrated good validity (CV < 0.25, r > 0.96) with the reference. Not reaching 100% accuracy is possibly due to saturation (clipping) in the acceleration signal. To keep count for this an IMU unit with a higher dynamic acceleration range or the application of a saturation compensating smoothing algorithm should be used. This could possibly lead to a more accurate classification system. The use of a mechanical limb with one rotation axis created extra undesired centripetal acceleration in the y-axis acceleration data of the IMU, resulting in higher calculated foot velocity. Usage of a mechanical limb with an extra element and rotation axis in its chain of motion is needed to minimalize the centripetal acceleration in the y-axis. The classification system in combination with the valid method for calculating foot velocity holds the potential to be used in field settings, assisting coaches and trainers with the performance analysis and load management of players.