Publications‎ > ‎2011‎ > ‎

Towards determining absolute velocity of freestyle swimming using 3-axis accelerometers

Reference:  A Stamm, DV Thiel, B Burkett, DA Jamesa, Towards determining absolute velocity of freestyle swimming using 3-axis accelerometers, Procedia Engineering 13, 120-125

Abstract:  Investigating the performance of an athlete and monitoring them is important to athletes and coaches. Coaches are not always on side when athletes doing their training, so a device which is small and easy to use will increase the monitored training sessions significantly and allow the athlete to compare multiple training sessions. In this research a small, portable inertial sensor platform was used to investigate the movement of swimmers and was set to record data at 100 Hz. The experiment was undertaken in an indoor pool with the sensor attached to the swimmer's sacrum, a velocity meter (Speed Probe 5000 – SP5000) attached to the swimmers suit with a video camera capturing the swimmer over the whole lap. The SP5000 measures the velocity directly and provides a synchronised video with the gathered velocity data. This system was used as main reference as it is already proven as a robust method and provides data files which can be directly imported into Matlab. The swimmer was asked to push-off with both feet against the wall and perform one freestyle stroke lap, which was repeated at different speeds. The timing parameters of the lap (i.e. start time, end time, stroke frequency) can be identified from the acceleration data. The acceleration data was then passed through a 0.5 Hz low pass filter to gain the sensor orientation, which was then removed for further processing. The velocity profile was calculated using the acceleration in swimming direction (ay) and the total acceleration (atot). The mean velocity from the SP5000 was 0.964 ± 0.086 m/s whereby the mean velocity derived from the accelerometer was 1.331 ± 0.207 m/s and 0.944 ± 0.119 m/s for ay and atot respectively. This research has shown that velocity information can be derived from acceleration data but there is still a difference in comparison to the SP5000 velocity. Future work needs to find a better approach in removing the sensor orientation.