TY - GEN
T1 - Accurate Horse Gait Event Estimation Using an Inertial Sensor Mounted on Different Body Locations
AU - Darbandi, Hamed
AU - Braganca, Filipe Serra
AU - van der Zwaag, Berend Jan
AU - Havinga, Paul
N1 - Publisher Copyright:
© 2022 IEEE.
PY - 2022
Y1 - 2022
N2 - Accurate calculation of temporal stride parameters is essential in horse gait analysis. A prerequisite for calculating these parameters is identifying the exact timings of gait events, i.e., hoof-on and hoof-off moments. A hoof-mounted inertial measurement unit (IMU) can be used to identify these moments accurately, yet this approach is often impractical due to the vulnerability of IMU to the impacts during locomotion. In this study, we investigated the possibility of accurately estimating the gait events using the signals of an IMU mounted on a less vulnerable location, such as a limb or upper body. To achieve the goal, we equipped IMUs on horses limbs, withers, and sacrum and measured them during different gaits. Then, we estimated the gait events timings by training recurrent neural networks models on the output signals of each IMU. Finally, we evaluated the models by comparing their results to the gait events timings labeled from hoof-mounted IMUs. The best performing model represented the best location (between the limbs, withers, and sacrum) for gait event estimation. Compared to the previous studies, our models yielded higher accuracy and were more generic by supporting more gaits. In conclusion, accurate calculation of temporal stride parameters is feasible by estimating gait event timings using an IMU mounted on less vulnerable body locations.
AB - Accurate calculation of temporal stride parameters is essential in horse gait analysis. A prerequisite for calculating these parameters is identifying the exact timings of gait events, i.e., hoof-on and hoof-off moments. A hoof-mounted inertial measurement unit (IMU) can be used to identify these moments accurately, yet this approach is often impractical due to the vulnerability of IMU to the impacts during locomotion. In this study, we investigated the possibility of accurately estimating the gait events using the signals of an IMU mounted on a less vulnerable location, such as a limb or upper body. To achieve the goal, we equipped IMUs on horses limbs, withers, and sacrum and measured them during different gaits. Then, we estimated the gait events timings by training recurrent neural networks models on the output signals of each IMU. Finally, we evaluated the models by comparing their results to the gait events timings labeled from hoof-mounted IMUs. The best performing model represented the best location (between the limbs, withers, and sacrum) for gait event estimation. Compared to the previous studies, our models yielded higher accuracy and were more generic by supporting more gaits. In conclusion, accurate calculation of temporal stride parameters is feasible by estimating gait event timings using an IMU mounted on less vulnerable body locations.
KW - Deep learning
KW - Gait
KW - Horse
KW - Inertial sensors
KW - 2023 OA procedure
UR - http://www.scopus.com/inward/record.url?scp=85136088094&partnerID=8YFLogxK
U2 - 10.1109/SMARTCOMP55677.2022.00076
DO - 10.1109/SMARTCOMP55677.2022.00076
M3 - Conference contribution
AN - SCOPUS:85136088094
SN - 978-1-6654-8153-3
T3 - Proceedings IEEE International Conference on Smart Computing, SMARTCOMP
SP - 329
EP - 335
BT - Proceedings - 2022 IEEE International Conference on Smart Computing, SMARTCOMP 2022
PB - IEEE
CY - Piscataway, NJ
T2 - 8th IEEE International Conference on Smart Computing, SMARTCOMP 2022
Y2 - 20 June 2022 through 24 June 2022
ER -