Sensors attached to the tendon boots of showjumping and dressage horses provided accurate data on training, opening the door to the use of the technology by coaches and riders.
Attaching sensors to the horse’s legs could give a simple solution that is accessible to all riders, they said. However, there is a scarcity of research on automatic classification of horse jumping and dressage training movements.
The Ghent University study team set out to use an advanced machine learning algorithm to categorize leg accelerometer data from typical dressage and jumping training motions.
They collected leg accelerometer data from 14 adult warmblood horses during jumping and dressage training at various Belgian stables. The horses ranged in their skill levels.
In all cases, an 11-gram six-axis accelerometer and data-logging device was attached to the side of each front tendon boot using Velcro and tape. One horse also had the sensors fitted to its hind tendon boots.
For the first time, six jumping training and 25 advanced horse dressage activities were classified using specifically developed models based on a neural network.
Jumping training could be automatically classified with an accuracy of 100%, while dressage training could be classified with an accuracy of 96.29%. Assigning the dressage action to classes results in higher accuracies of 98.87%, 99.10% and 100%, respectively.
During dressage training, side movement could be identified with an accuracy of 97.08%.
A velocity estimation model was also developed, based on the measured speeds of seven horses performing the collected, working, and extended gaits during dressage training. It was shown that the horses’ velocities can be correctly estimated, resulting in a stride length estimation model.
“These results can help in further development of an automatic system for training activity detection and help improve training sessions and horses’ fitness levels,” the authors concluded.
They predicted that, with more training data from different breeds, their system would be even more robust.
The Ghent study team comprised Eerdekens, Margot Deruyck, Jaron Fontaine, Bert Damiaans, Luc Martens, Eli De Poorter, Jan Govaere, David Plets and Wout Joseph.
Eerdekens, A.; Deruyck, M.; Fontaine, J.; Damiaans, B.; Martens, L.; De Poorter, E.; Govaere, J.; Plets, D.; Joseph, W. Horse Jumping and Dressage Training Activity Detection Using Accelerometer Data. Animals 2021, 11, 2904. https://doi.org/10.3390/ani11102904