The use of wearable sensors for animal behaviour assessment

Abstract

The research outlined in this thesis presents novel applications of wearable sensors in the domain of animal behaviour assessment. The use of wearable sensing technology, and in particular accelerometry, has become a mainstay of behaviour assessment in humans, allowing for detailed analysis of movement based behaviour and health monitoring. In this thesis we look to apply these methodologies to animals and identify approaches towards monitoring their health and wellbeing. We investigate the use of the technology in the animal domain through a series of studies examining the problem across multiple species and in increasingly complex scenarios. A tightly constrained scenario is presented initially, in which horse behaviour was classi ed and assessed in the context of dressage performances. The assessment of lying behaviour in periparturient sows con ned to gestation crates examines a scenario in which the movement of the subject was constrained, but not predetermined. Expanding this work to include sows housed in free-farrowing environments removed the movement constraints imposed by the gestation crates. We examine the implications of the use of multiple sensors and how this might a ect the accuracy of the assessments. Finally, a system for behaviour recognition and assessment was developed for domestic cats. Study animals were free to move and behave at their own discretion whilst being monitored through the use of wearable sensors, in the least constrained of the studies. The scenarios outlined herein describe applications with an increasing level of complexity through the removal of constraints. Through this work we demonstrate that these techniques are applicable across species and hold value for the wellbeing of both commercial and companion animals.