Detecting Equine Gaits Through Rider-Worn Accelerometers

Automatic horse gait classification offers insights into training intensity, but direct<br>sensor attachment to horses raises concerns about discomfort, behavioral disruption, and<br>entanglement risks. To address this, our study leverages rider-centric accelerometers for<br>moveme...

Full description

Saved in:
Bibliographic Details
Main Authors: Jorn Schampheleer, Anniek Eerdekens, Wout Joseph, Luc Martens, Margot Deruyck
Format: Article
Language:English
Published: MDPI AG 2025-04-01
Series:Animals
Subjects:
Online Access:https://www.mdpi.com/2076-2615/15/8/1080
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849712325182881792
author Jorn Schampheleer
Anniek Eerdekens
Wout Joseph
Luc Martens
Margot Deruyck
author_facet Jorn Schampheleer
Anniek Eerdekens
Wout Joseph
Luc Martens
Margot Deruyck
author_sort Jorn Schampheleer
collection DOAJ
description Automatic horse gait classification offers insights into training intensity, but direct<br>sensor attachment to horses raises concerns about discomfort, behavioral disruption, and<br>entanglement risks. To address this, our study leverages rider-centric accelerometers for<br>movement classification. The position of a sensor, sampling frequency, and window size of<br>segmented signal data have a major impact on classification accuracy in activity recognition.<br>Yet, there are no studies that have evaluated the effect of all these factors simultaneously<br>using accelerometer data from four distinct rider locations (the knee, backbone, chest, and<br>arm) across five riders and seven horses performing three gaits. A total of eight models<br>were compared, and an LSTM-convolutional network (ConvLSTM2D) achieved the highest<br>accuracy, with an average accuracy of 89.72% considering four movements (halt, walk,<br>trot, and canter). The model performed best with an interval width of four seconds and<br>a sampling frequency of 25 Hz. Additionally, an F1-score of 86.18% was achieved and<br>validated using LOSOCV (Leave One Subject Out Cross-Validation).
format Article
id doaj-art-1954d0d154ab4e3390708ec73d737baa
institution DOAJ
issn 2076-2615
language English
publishDate 2025-04-01
publisher MDPI AG
record_format Article
series Animals
spelling doaj-art-1954d0d154ab4e3390708ec73d737baa2025-08-20T03:14:19ZengMDPI AGAnimals2076-26152025-04-01158108010.3390/ani15081080Detecting Equine Gaits Through Rider-Worn AccelerometersJorn Schampheleer0Anniek Eerdekens1Wout Joseph2Luc Martens3Margot Deruyck4WAVES, Department of Information Technology, Ghent University-IMEC, 9052 Ghent, BelgiumWAVES, Department of Information Technology, Ghent University-IMEC, 9052 Ghent, BelgiumWAVES, Department of Information Technology, Ghent University-IMEC, 9052 Ghent, BelgiumWAVES, Department of Information Technology, Ghent University-IMEC, 9052 Ghent, BelgiumWAVES, Department of Information Technology, Ghent University-IMEC, 9052 Ghent, BelgiumAutomatic horse gait classification offers insights into training intensity, but direct<br>sensor attachment to horses raises concerns about discomfort, behavioral disruption, and<br>entanglement risks. To address this, our study leverages rider-centric accelerometers for<br>movement classification. The position of a sensor, sampling frequency, and window size of<br>segmented signal data have a major impact on classification accuracy in activity recognition.<br>Yet, there are no studies that have evaluated the effect of all these factors simultaneously<br>using accelerometer data from four distinct rider locations (the knee, backbone, chest, and<br>arm) across five riders and seven horses performing three gaits. A total of eight models<br>were compared, and an LSTM-convolutional network (ConvLSTM2D) achieved the highest<br>accuracy, with an average accuracy of 89.72% considering four movements (halt, walk,<br>trot, and canter). The model performed best with an interval width of four seconds and<br>a sampling frequency of 25 Hz. Additionally, an F1-score of 86.18% was achieved and<br>validated using LOSOCV (Leave One Subject Out Cross-Validation).https://www.mdpi.com/2076-2615/15/8/1080equinesaccelerometer sensoranimal activity recognitionmachine learningconvolutional LSTM network
spellingShingle Jorn Schampheleer
Anniek Eerdekens
Wout Joseph
Luc Martens
Margot Deruyck
Detecting Equine Gaits Through Rider-Worn Accelerometers
Animals
equines
accelerometer sensor
animal activity recognition
machine learning
convolutional LSTM network
title Detecting Equine Gaits Through Rider-Worn Accelerometers
title_full Detecting Equine Gaits Through Rider-Worn Accelerometers
title_fullStr Detecting Equine Gaits Through Rider-Worn Accelerometers
title_full_unstemmed Detecting Equine Gaits Through Rider-Worn Accelerometers
title_short Detecting Equine Gaits Through Rider-Worn Accelerometers
title_sort detecting equine gaits through rider worn accelerometers
topic equines
accelerometer sensor
animal activity recognition
machine learning
convolutional LSTM network
url https://www.mdpi.com/2076-2615/15/8/1080
work_keys_str_mv AT jornschampheleer detectingequinegaitsthroughriderwornaccelerometers
AT anniekeerdekens detectingequinegaitsthroughriderwornaccelerometers
AT woutjoseph detectingequinegaitsthroughriderwornaccelerometers
AT lucmartens detectingequinegaitsthroughriderwornaccelerometers
AT margotderuyck detectingequinegaitsthroughriderwornaccelerometers