Arctic HARE. A Machine Learning-based System for Performance Analysis of Cross-country Skiers

Abstract

The advances in sensor technology and big-data processing enable performance analysis of sport athletes. With the increase in data, both from on-body sensors and cameras, it is possible to quantify what makes a good athlete. However, typical approaches in sports performance analysis are not adequately equipped for automatically handling big data.

This thesis presents Arctic Human Activity Recognition on the Edge, a machine-learning based system that aims to provide live performance analysis of cross-country skiers. Arctic HARE uses on-body sensors and cameras to capture movement of the skier, and provides classification of the perceived technique. We explore and compare two approaches to classifying data, in order to determine optimal representations that embody the movement of the skier.

The viability of Arctic HARE is substantiated through a working prototype. We ascertain how to optimally capture the movement of the skier and we qualitatively compare the two approaches through experimental evaluation. Our results reveal we can achieve as high as 97% accuracy for real-time classification of cross-country techniques.