| Title |
Analysis of action recognition using skeleton data / |
| Translation of Title |
Veiksmų atpažinimo analizė naudojant skeleto duomenis. |
| Authors |
Vaitiekus, Hermanas |
| Full Text |
|
| Pages |
35 |
| Keywords [eng] |
Video classification, One-Class clasification, Snatch lift, CrossFit, Movement analysis, Skeleton data, Dynamic Time Warping (DTW), Outlier detection |
| Abstract [eng] |
This thesis focused on one-class classification of actions video records. The thesis covers developing a framework for analyzing snatch lifts in CrossFit, utilizing skeleton data and Dynamic Time Warping (DTW) to assess the quality of the lift. The objective was to provide a data driven evaluation of the movement, complementing traditional coaching methods. The study began with skeleton data extraction, followed by data normalization techniques, including scaling, centering, rotation, and smoothing. A golden standard was created and the performance of other lifts was compared against this standard using DTW. Outliers were detected to identify lifts with significant deviations from optimal form. The results demonstrated that the proposed framework identifies deviations from the golden standard, offering a promising tool. This study contributes to the growing body of research in sports performance analysis, particularly in the context of CrossFit. |
| Dissertation Institution |
Vilniaus universitetas. |
| Type |
Master thesis |
| Language |
English |
| Publication date |
2025 |
