INTRODUCTION: The ideal cross-country ski training on a treadmill aims to best simulate conditions of an outdoor competition or exercise session. To reach this aim a self-adjust treadmill speed system depending on athlete position has been developed, thereby rendering indoor training even more real. The new system is based on a Microsoft Kinect device used as sensor. METHODS: Kinect sensor (Range: 1 to 4 m, Resolution: 1 mm, Field of view: 57° horizontal - 43° vertical) defines the entire body by recognizing 20 skeleton joints. In particular, in this application the hip centre was used to determine athlete position as it roughly represents the centre of mass. The Kinect was positioned in front of an oversize treadmill (RL3500E, Rodby, Sweden) to optimal distance to sense athlete's entire body allowing for the best recognition. Treadmill length was segmented in three zones (neutral, frontal and posterior) where athlete could be detected. The acquired athlete's position are processed by the function build to control treadmill speed. Skier positioned inside neutral zone does not lead to any speed change, while skier positioned in the frontal or in the posterior zones drives to respectively increase and decrease treadmill speed. Non linear penalty function has been adopted in the frontal and posterior zones to weight on displacements detected. Parameters describing the PID controller can be tuned on the athlete preference and ski technique. RESULTS: Kinect sensor was able to detect any position change with a good accuracy (compare to optical motion capture system as tested with Pro Reflex Qualisys system with a BlandAltman correlation r = 0.98).The system was tested for different cross-country skiing classical techniques and proved to be able to efficiently change the speed accordingly to skiers position and therefore to be a good tool for self-selected speed. DISCUSSION: Many self-adjusting speed systems for treadmill have been developed so far; nevertheless, a system based on Kinect technology is innovative. Compared to a laser sensor, Kinect is able to detect athlete position even in case of unbalance or lateral movements, thanks to its wide field of view. External objects like ski poles are not interfering with skeleton tracking, thereby allowing a robust position detection. In addition, Kinect sensor has the advantage of being wireless and relatively cheap. CONCLUSION: The novelty of using Microsoft Kinect permits an accurate and robust detection of the athlete. Being the speed dependent on athlete physical conditions, elevation, skiing techniques and distance travelled, the self-adjusting speed system is able to better simulate real condition of outdoor training. In addition, thanks to sensor skills to detect segments body position, the system permits to implement solutions to acquire, display in real time and store some parameters related to athlete locomotion such as cycle frequency or articular movements. --- The hosting University of Jyväskylä is planning to publish conference proceedings "Science and Nordic Skiing III". In case you are interested in this publication please contact the editors (Anni Hakkarainen anni.s.j.hakkarainen@jyu.fi) to become registered for the book.
© Copyright 2015 3rd International Congress on Science and Nordic Skiing - ICSNS 2015. 5-8 June 2015, Vuokatti, Finland. Veröffentlicht von University of Jyväskylä; University of Salzburg. Alle Rechte vorbehalten.