A Machine Learning Approach to Analyze Rider's Effects on Horse Gait Using On-Body Inertial Sensors

Wearable sensors play an essential role not only in human gait analysis but also in equine gait studies. One of their roles is evaluating the rider's effects on horse biomechanics, which affects the horse's well-being and performance. Therefore, it is necessary to identify whether the horse is being ridden or not before performing gait analysis, particularly in applications that require automatic and real-time evaluation. However, there is a lack of study to automatically identify whether the horse is being ridden or not ridden despite several studies that developed machine learning models for labeling various gait characteristics. This study investigated the possibility of classifying horse locomotion to ridden/not ridden labels using minimal inertial sensors on the horse's body. We presented a classification model that accurately labels the horse's ridden state using a minimum number of wearable sensors.