| Summary: | The growing interest in skateboarding as a competitive sport requires new motion analysis approaches and innovative ways to portray athletes’ results as previous techniques in the identification of the tricks was often inadequate in providing accurate evaluation during competition. Therefore, there is a need to introduce an unprejudiced method of evaluation in skateboarding competitions. This paper presents the classification of five different skateboarding tricks (Ollie, Kickflip, Frontside 180, Pop Shove-it, and Nollie Frontside Shove-it) through the identification os significant frequency-domain signals collected via Inertial Measurement Unit (IMU) and the use of machine learning models. Onemale skateboarder (age: 23 years old) performed five different tricks repeatedly for several times. The time-domain data acquired from the IMU were converted to frequency-domain by employing Fast Fourier Transform (FFT) and a number of statistical features (mean, kurtosis, skewness, standard deviation, root mean square and peak-to-peak corresponding to x-y-z-axis of the IMU) were then extracted. Significant features were then identified from the Information Gain (IG) scoring. It was shown from the study that the Naïve Bayes (NB) classifier is able to acquire the highest classification accuracy of 100% on the test data compared to the other evaluated classifiers, namely Artificial Neural Network (ANN) and SupportVector Machine (SVM), by utilising the selected features, suggesting that the proposed methodology could provide an objective-based evaluation of the tricks
|
|---|