The popularity of skateboarding continuously grows for athletes performing the sport and for spectators following competitions. The presentation and the assessment of the athletes' performance can be supported by state-of-the-art motion analysis and pattern recognition methods. In this paper, we present a trick classi cation analysis based on motion data of inertial measurement units. Six tricks were performed by seven skateboarders. A trick event detection algorithm and four di erent classi cation methods were applied to the collected data. A sensitivity of the event detection of 94.2% was achieved. The classi cation of correctly detected trick events provides an accuracy of 97.8% for the best performing classi ers. The proposed algorithm holds the potential to be extended to a real-time application that could be used to make competitions fairer, to better present the assessment to spectators and to support the training of athletes.
© Copyright 2015 KDD Workshop 9: Workshop on Large-Scale Sports Analytics. All rights reserved.