The Classification of Skateboarding Trick Manoeuvres: A K-Nearest Neighbour Approach

The evaluation of skateboarding tricks is commonly carried out subjectively through the prior experience of the panel of judges during skateboarding competitions. Hence, this technique evaluation is often impartial to a certain degree. This study aims at classifying flat ground tricks namely Ollie, Kickflip, Shove-it, Nollie and Frontside 180 through the use of Inertial Measurement Unit (IMU) and a class of machine learning model namely k-Nearest Neighbour (k-NN). An amateur skateboarder (23 years of age ± 5.0 years’ experience) executed five tricks for each type of trick repeatedly on a customized ORY skateboard (IMU sensor fused) on a cemented ground. A number of features were extracted and engineered from the IMU data, i.e., mean, skewness, kurtosis, peak to peak, root mean square as well as standard deviation of the acceleration and angular velocities along the primary axes. A variation of k-NN algorithms were tested based on the number of neighbours, as well as the weight and the type of distance metric used. It was shown from the present preliminary investigation, that the k-NN model which employs k = 1 with an equal weight applied to the Euclidean distance metric yielded a classification accuracy of 85%. Therefore, it could be concluded that the proposed method is able to classify the skateboard tricks reasonably well and will in turn, assist the judges in providing more accurate evaluation of the tricks as opposed to the conventional-subjective based assessment that is applied at present.