Abstract des Autors

High precision Global Navigation Satellite System (GNSS) measurements are becoming more and more popular also in Alpine skiing due to relatively undemanding setup and excellent performance. Unfortunately, GNSS basically provides only a single point measurement, where the antenna is placed, typically behind the skier's neck. The key issue is how to estimate other even more relevant parameters, like center of mass (COM) trajectories, skis trajectories, and skier's pose. One solution to the above problem was proposed in Supej, Kugovnik, and Nemec (2008), where the skier's body was approximated with a quasi-statically balanced inverted pendulum. Kalman filtering was used to estimate accelerations necessary to compute the inverted pendulum equilibrium pose. Similar approach was proposed also in Gilgien et al. (2013). The inverted pendulum approach was enhanced, taking into account also the inverted pendulum dynamics where it was used to study energy dissipation caused by air drag in giant slalom skiing (Supej et al., 2013). However, approximation of the skier's pose using only a "bar" is very rough; therefore it is of great importance weather we can obtain better approximation using a multi segment skeleton model. In this study we propose a solution based on generalization of skiing using back propagation neural networks (NN) and compare it with inverted pendulum based method. The reference measurement setup used for validation of both methods was a setup composed of inertial motion capture and GNSS measurement, respectively (Supej, 2010).