Author(s): FESER, E., LINDLEY, K., PERRY, L., STUCKER, R., GILLARD, C., Institution: BELMONT UNIVERSITY, Country: UNITED STATES, Abstract-ID: 1421

INTRODUCTION:
Researchers have determined that a single, sacrum-placed inertial measurement unit (IMU) sensor can be used to record pelvis motion during sprinting [1]. This approach is convenient but lacks reference to other valuable metrics needed to track pelvis motion during sprinting (e.g. contact detection). Signal processing techniques can be used to find step kinematic metrics [2,3] but have not been evaluated for short, accelerated sprints. The purpose of this study was to establish the magnitude of systematic bias and random error in determining ground contact time (CT) obtained from a sport market ready IMU to those obtained from a high-speed camera recording during a 20 m sprint run.
METHODS:
Seventeen competitive collegiate athletes completed a 20 m sprint with an IMU and two high-speed video cameras (iPhone 11, Apple Inc., USA; frame rate = 240 Hz) simultaneously recording. The IMU (Blue Trident, Vicon Motion Systems, UK) was placed on the posterior sacrum. To process the IMU data, acceleration vectors were rotated into the global reference frame using the on-board global orientation estimates in quaternion format. This resulted in a vertical acceleration value represented by the third element of the rotated acceleration vector, which was filtered by a fourth order low-pass butterworth filter with a 10 Hz cutoff frequency. CT for steps 1, 2, 4, 7, and 10 were identified as the time between the maximum and minimum vertical acceleration peaks following methods described in [4]. The video recordings were manually analyzed (Kinovea, vers. 0.9.5). To compare CT derived from the IMU to the cameras the following, bias (mean measurement difference between the two devices, IMU – Camera) and random error (1.96 × standard deviation of the differences between the devices) were determined [5].
RESULTS:
An average negative bias, indicating a lower measurement for the IMU, was found for all steps (1: -67 ms; 2: -45 ms; 4: -33 ms; 7: -28 ms; 10: -3 ms). Bias measures reported as a percent of the CT identified by video are -35%, -28%, -25%, -22%, and -2.5% for steps 1, 2, 4, 7, and 10, respectively. Random error was largest for step 1 (±95 ms). For steps 2, 4, 7, and 10 random error was ±90 ms, ±76 ms, ±68 ms, and ±78 ms respectively.
CONCLUSION:
This study evaluated the use of a sacrum mounted IMU to determine CT for individual steps during a 20 m accelerated sprint. An average negative bias was found for all steps, indicating a lower CT time for the IMU method. The bias values lessened as the sprint progressed, reaching -3 ms at step 10. However, random error at step 10 (±78 ms) represented ±35% of the reference video mean CT. Although bias was relatively low by step 10, practitioners should consider if the potential for error is acceptable with respect to their specific application context.
1. Wada et al. (2020) 2. Day et al. (2021) 3. Lee et al. (2010) 4. Miranda-Oliveira et al. (2023) 5. Bland & Altman (1999)