The classification of skateboard trick manoeuvres through the integration of inertial measurement unit (imu) and machine learning
The growing interest in skateboarding as a competitive sport requires new motion analysis approaches and innovative ways to portray athletes' performance as previous classification of tricks techniques was often deemed inadequate in providing accurate evaluation during competition. Therefore, a...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2022
|
| Subjects: | |
| Online Access: | http://umpir.ump.edu.my/id/eprint/37630/1/ir.The%20classification%20of%20skateboard%20trick%20manoeuvres%20through%20the%20integration%20of%20inertial%20measurement%20unit%20%28imu%29%20and%20machine%20learning.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | The growing interest in skateboarding as a competitive sport requires new motion analysis approaches and innovative ways to portray athletes' performance as previous classification of tricks techniques was often deemed inadequate in providing accurate evaluation during competition. Therefore, an objective and fair means of evaluating skateboarding tricks were developed to analyze skateboarder’s tricks is non-trivial. This study aims at classifying flat ground tricks, namely Ollie, Kickflip, Pop Shove-it, Nollie Frontside Shove-it, and Frontside 180, through the use of Inertial Measurement Unit (IMU) and machine learning models. Six armature skateboarders executed five tricks for each type of trick repeatedly by five times. It is worth noting that the time-series (TS) domain input skateboard data were transformed to two different types of frequency domains, namely Fast Fourier Transform (FFT) and Discrete Wavelet Transform (DWT). Therefore, both the time and frequency domain features were used to evaluate six machine learning models, Logistic Regression (LR), Random Forest (RF), k-Nearest Neighbors (k-NN), Artificial Neural Network (ANN), Naïve Bayes (NB), and Support Vector Machine (SVM). In addition, two types of feature selection methods known as Wrapper and Embedded methods, were applied to identify the significant features. The datasets were split into 70:30 ratios for training and testing, respectively. It was shown from the study, that the RF-TS (All), RF-TS (Wrapper), RF-TS (Embedded), RF-DWT (All), RF-DWT (Wrapper), and RF-DWT (Embedded) yield 100% classification accuracy. Nevertheless, the RF-TS (Wrapper) is established to be the best as it utilises the least number of features (forty-one instead of fifty-four), which in turn reduces the complexity of the model for the classification of the tricks evaluated. Therefore, it is opined that the approach proposed can reasonably identify the tricks of the skateboard to help the judges evaluates the trick performances more precisely as opposed to the currently used subjective and traditional techniques. |
|---|