Area-specific vocalisation and activity monitoring in pigs

Vocalisations and activity are important factors for monitoring animal behaviour. Previous studies have shown that these parameters are altered in the context of disease, farrowing and distress. In the highly standardized conditions of livestock farming, these contexts are often closely linked to specific areas of the barn – allowing information about specific animal states to be inferred from these rather unspecific parameters. We have developed a sensor that uses acoustic beamforming and radar to measure vocalisations and activity in defined areas. The sensor was tested in two settings. Firstly, recently weaned piglets were socially isolated with unrestricted sensory contact for up to 20 minutes. During this comparatively mild distress, stress vocalisations could be detected with 63 % sensitivity, 97 % specificity, and 79 % balanced normalized precision. Secondly, data recorded in the context of farrowing demonstrated that the sensor is able to distinguish between movement and vocalisation events in adjacent pens. To quantify the sensor’s spatial differentiation, we compared an empty and an adjacent occupied farrowing pen over the time of birth, using Receiver Operating Characteristics curve (ROC). The ROC curves for discriminating the activity in the pens reached an area under the curve (AUC) of 0.87 to 0.95, whereby 0.5 would be the worst and 1.0 the best performance. The ROC curves for discriminating sources of stress events reached an AUC of 0.86 to 0.96. Furthermore, this technology is currently being adapted for applications in other farm animal species.