Automatic Recognition of Motor Skills in Triathlon
Author Information
Author(s): Chesher Stuart M., Martinotti Carlo, Chapman Dale W., Rosalie Simon M., Charlton Paula C., Netto Kevin J.
Primary Institution: Curtin University
Hypothesis
Can a peak detection algorithm and machine learning model effectively measure movement cadence and recognize cycling task changes in triathletes?
Conclusion
The study found that a custom peak detection algorithm and machine learning model are effective tools for automatically analyzing triathlon performance.
Supporting Evidence
- The peak-counting algorithm detected 98.7% of swimming strokes and 97.8% of cycling pedal strokes.
- The machine learning model was 94% accurate at distinguishing between 'in-saddle' and 'out-of-saddle' riding.
- Participants were filmed while competing to ensure accurate data collection.
Takeaway
This study created a tool that helps triathletes measure their swimming, cycling, and running movements automatically, making it easier to improve their performance.
Methodology
Triathletes wore a trunk-mounted measurement unit while competing in a triathlon, and their movements were recorded and analyzed using a peak detection algorithm and machine learning model.
Potential Biases
The initial data collection had an overrepresentation of 'in-saddle' riding, which may have inflated the model's accuracy.
Limitations
The model could not distinguish between 'coasting' and 'in-saddle' riding, limiting the analysis of cycling motor skills.
Participant Demographics
12 triathletes (6 males, 6 females; mean age 16.2 years) with varying competition levels.
Statistical Information
P-Value
p<0.001
Statistical Significance
p<0.001
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website