Estimating ski orientation using IMUs in alpine skiing

Abstract

Ski-snow interaction is the essential component of alpine skiing. To understand how a skier manipulates his ski to turn, we need to develop methods to measure the orientation of the ski throughout a complete run. Recent studies tried to use IMUs to estimate edge angle (EA) during skiing. We introduce and validate a method on how to calibrate and employ IMUs to precisely and accurately measure roll angles (RA) as a matter of changing orientation of the ski around its longitudinal axis in 3D space during skiing. Static orientation measurements on an inclined plane perfectly correlate (r2 = 1) with 3D motion capturing: RMSE = 0.18° and 0.24° respectively. Bland Altman showed a mean bias of 0.23° (95% CI: -0.16°, 0.63°) and 0.21° (95% CI: -0.3°, 0.73°). Accuracy and drift tests against constant standardised rotational velocities showed no drift behaviour over time, but RA estimation accuracy is reduced with increasing angular velocities (SD @ ±300°/s: 0.57°, max. difference from average at ±300°/s: 2.7°). During skiing on a ski ergometer the comparison of maximum RA against Vicon showed a mean bias of 0.13° (95% CI: -0.86° to 1.1°). Even though ski ergometer skiing has a similar frequency and angular velocity profile like outdoor skiing, there are more rotational degrees of freedom in outdoor skiing. The foundation is provided in this paper. To understand how a skier manipulates the ski on snow and to understand RA and EA progression during a turn in detail, further research should validate the method in the field and additionally look into RA progression within individual turns.

Keywords

• edge angle
• roll angle
• ski-snow interaction
• inertial measurement units
• validation
• Madgwick filter