Gait analysis in horse sports

In modern showjumping and cross-country riding, horse-rider-pairs have to jump a series of obstacles in a given time. A jump is considered successful (penalty-free) if a horse can comfortably jump the fence without elements of the fence falling down. If any of the elements of the fence falls down or the horse refuses to jump, the rider obtains penalty points or can be disqualified from the competition. An unsuccessful jump can lead to injury and loss in trust of the rider. The success of a jump is determined by the number, length and harmony of strides a horse performs before a fence. We propose a system for tracking horse strides and jumps using a smartphone attached to the horse's saddle. Our system detects and segments individual strides and computes the length of a stride using signal processing and machine learning methods. We collected data from 9 horses who performed several jumps. Our results indicate that our system can detect horse strides with a precision of 96.3%, a recall of 95.7% and a pearson correlation of 0.73 with respect to our ground truth data set. We further describe a method to characterise the canter gait of the horse. Our system is intended to be used by riders to adapt their training and competition strategies to the physical limitations of the horse. The rider can thus prevent accidents due to an overtaxing of the horse or miscalculation of canter strides by the rider.