Development of a skateboarding trick classifier using accelerometry and machine learning

Abstract

Skateboarding is one of the most popular cultures in Brazil, with more than 8.5 million skateboarders. Nowadays, the discipline of street skating has gained recognition among other more classical sports and awaits its debut at the Tokyo 2020 Summer Olympic Games. This study aimed to explore the state-of-the-art in inertial measurement unit (IMU) use for skateboarding trick detection and to develop new classification methods using supervised machine learning and artificial neural networks (ANN). State-of-the-art knowledge on motion detection in skateboarding was used to generate 543 artificial acceleration signals via signal modeling, corresponding to 181 flat-ground tricks divided into five classes (NOLLIE, NSHOV, FLIP, SHOV, OLLIE). The classifier consisted of a multilayer feed-forward neural network created with three layers and a supervised learning algorithm (backpropagation). The use of ANNs explicitly trained for each measured acceleration axis yielded error percentages below 0.05%, with computational efficiency enabling real-time operation. Machine learning can be a helpful technique for classifying flat-ground skateboarding tricks, assuming that the classifiers are properly constructed and trained and that the acceleration signals are preprocessed correctly.

BibTeX

@article{correa2017development,
  title={Development of a skateboarding trick classifier using accelerometry and machine learning},
  author={Corr{\^e}a, Nicholas Kluge and Lima, J{\'u}lio C{\'e}sar Marques de and Russomano, Thais and Santos, Marlise Araujo dos},
  journal={Research on Biomedical Engineering},
  volume={33},
  pages={362--369},
  year={2017},
  publisher={SciELO Brasil}
}