Applying common filtering processes to Global Navigation Satellite System-derived acceleration during team sport locomotion

This study aimed to observe whether there were substantial differences in acceleration during team-sport locomotion between GNSS manufacturers. Speed and acceleration were obtained from 7 professional rugby league athletes via 2 GNSS manufacturers (GPSports EVO, 10 Hz and STATSports Apex, 10 Hz) worn together during the same training sessions (n = 13). Raw GNSS data were exported from respective proprietary software and a 1 Hz, 4th-order Butterworth filter applied, with differences in speed and acceleration calculated between manufacturers. To determine the difference in acceleration and speed, a root mean square deviation (RMSD) was used. Linear mixed models were used and no substantial differences were found between manufacturers in raw and filtered speed variables. RMSD for average acceleration (m · s-2) decreased from raw (RMSD: 1.77 ± 0.37 m · s-2) to those seen at the filtered (RMSD: 0.27 ± 0.23 m · s-2) and twice filtered (0.24 ± 0.23 m · s-2) variables. Raw average acceleration (m · s-2) was substantially higher in Apex compared to EVO (Difference (Diff); CI: -0.82; -0.84 to -0.80). Following application of the common filter there was no substantial difference between GNSS models for average acceleration (Diff; CI: -0.04; -0.04 to -0.04). Acceleration variables derived from each manufacturer's proprietary software were substantially different.