Edge angle estimation using IMUs - An in-field validation apporach using AI pose estimation

(Schätzung von Kantenwinkeln mithilfe von IMUs - Ein Ansatz zur Validierung unter realen Bedingungen unter Verwendung einer KI-basierten Schätzung der Körperhaltung)

INTRODUCTION: Ski-snow interaction is the essential component of alpine skiing. A skier changes his trajectory and speed by manipulating the orientation and loading pattern, generating a reaction force from the snow surface which results in a change of direction (Reid et al., 2020). Using IMUs to determine, edging angles allows to measure the ski orientation throughout a complete run, but a concern is sensor offset, sensor drift and saturation leading to wrong estimations of the ski orientation. In simulating skiing with similar angular velocity and cycle frequencies we were able to estimate peak edge angles within ±1.96° (95% LoA) of a reference system using IMUs (Hummel et al., 2024) applying the Madgwick filter (Madgewick et al., 2011). In this study, we want to validate not just peak angles, but also investigate curve progression in an on-snow experiment. METHODS: Two IMUs (1000 Hz, 6-DOF, 16-bit, ± 2000°/s, ± 16 g) are mounted in two different setups: 1) To the back of both ski boots. 2) To the top of the left Ski and to the back of the left ski boot. Before IMUs are mounted, axes are adjusted and the accelerometer and gGyroscope are calibrated. Three trials each are performed by former ski racers and coaches (n = 3). Every trial starts with a five-second zero-movement-phase to remove gyroscope offset and retrieve the initial orientation. Trials consist of a full hill run, of which the edge angle will be estimated throughout. The last section of each trial is evaluated against a 3D AI pose estimation system (Nemo, Simi Reality Motion Systems GmbH ©, Unterschleißheim, Germany). Calf segment orientation is compared with regards to angle progression and peak angle against the estimation of the IMUs for setup 1. Further, the deviation between ski and boot orientation is evaluated with setup 2. To assess the overall quality of the AI System`s calibration a differential GPS is added to the ski and the 3D speed is compared between both systems. RESULTS/DISCUSSION: Pilot tests and the previous experiments on snow suggest plausible results, where no drift or saturation was detected. The accuracy and precision of the angels are yet to be investigated and results will be presented at the conference. CONCLUSION: Overall, the Madgwick filter is a promising approach to measure accurate edging angels for alpine skiing. The combination with pressure insoles and ski position is subject of further research to determine reaction forces and therefore, effects on change in the ski´s trajectory.
© Copyright 2025 10th International Congress on Science and Skiing, January 28 - February 1, 2025, Val di Fiemme, Italy. Alle Rechte vorbehalten.

Bibliographische Detailangaben
Schlagworte:
Notationen:Kraft-Schnellkraft-Sportarten
Tagging:Validität künstliche Intelligenz
Veröffentlicht in:10th International Congress on Science and Skiing, January 28 - February 1, 2025, Val di Fiemme, Italy
Sprache:Englisch
Veröffentlicht: 2025
Seiten:75
Dokumentenarten:Kongressband, Tagungsbericht
Level:hoch