Automatic Event Detection Using Wearable Technology During Short-Track Speed Skating Races

The performance in short-track speed skating (STSS) is driven by technique optimization. However, because of discomfort, clutter, or complexity, regular instrumentation may not be suitable for use in daily training. The objective of this study was to validate a single-accelerometer-based algorithm: 1) to detect the number of strokes and 2) accurately classify left, right, pivot, and straight-line strokes during four- and nine-lap practice race simulations. Twenty-eight athletes from the Canadian National STSS team were instrumented with an accelerometer taped to their sacrums that would collect tridimensional accelerations and angles from start to finish, and they were filmed with a single camera setup during four-lap and/or nine-lap individual race trials. Data were analyzed with a custom MATLAB algorithm and compared to video data on two datasets to investigate the number of strokes, pivots, and straights detected. Over 98% of strokes were detected; and over 99% right/left strokes, 97.7% pivots, and 98.6% straights were identified. The validation led to intraclass correlation coefficients [ICC(3, 1)] of over 0.97, indicating an excellent agreement between the two methods. The results support the ability of wearable technology to deliver valid speed-skating data, enabling rapid feedback to coaches and athletes with minimal equipment in training.