CENTRE OF MASS TRAJECTORY IN SNOWBOARD GIANT SLALOM USING INERTIAL SENSORS: LABORATORY AND IN-FIELD PRELIMINARY EVALUATION

The purpose of the present study was to evaluate the reconstruction accuracy of the centre of mass during snowboard giant slalom using inertial sensors (Opal, APDM, 128 Hz). Two approaches were implemented and tested: i) a multi-segment model using 7 inertial sensors on the trunk, the pelvis, the thighs, the shanks, and the board; and ii) a double integration of the acceleration at L5 level measured with one inertial sensor. The accuracy of the algorithms was verified in two laboratory conditions: a) the multi-segment model approach was tested indoor during controlled movements using stereo-photogrammetry as gold standard, and b) the double integration of acceleration approach was tested outdoor in simulated movements on a longboard using GPS as gold standard. Successively, to verify the application in real conditions, an in-field acquisition of a forerunner athlete during a snowboard world cup competition was performed. The position of the centre of mass estimated indoor with multi-segmental model approach reported in the local reference frame of the board showed high correlation with respect to stereo-photogrammetry (r=0.87) and a RMS error of 3.8 [%] expressed as percentage of the range of motion during the trial (1.32m). For the simulated movements test in outdoor conditions on the longboard applying the double integration approach, high correlation was found with respect to the GPS data (r=0.95) on the trajectory but , for the 4 turns trial, a RMS difference on the distance equal to 15.3 [%] expressed as percentage of the whole distance covered (46m). Finally, the in-field acquisition showed how using inertial sensors is a viable option for collecting centre of mass data during training session useful for coaches and athletes. The approach using one sensors at L5 level showed low level of accuracy with respect to the one using a multi-segment model. Further developments should be performed in the direction of a better estimation of the orientation of the inertial sensors and of the boundary conditions for the integration algorithm.