Auditory recordings at pen level in farrowing units have potential for health and welfare monitoring but are limited by the close proximity of neighbouring animals. This pilot study assessed whether vocalizations recorded with a single microphone placed above one pen differed acoustically depending on whether they originated from inside or outside the focal pen. A total of 475 sow and piglet vocalizations were manually selected and labelled using synchronized video recordings. Principal Component Analyses (PCA) showed no clear differentiation between in-pen and out-of-pen vocalizations for grunts or squeals, despite explaining over 85% of the total variance. In contrast, PCA restricted to piglet screams revealed partial separation along the first principal component (t-test, p < 0.01). A complementary logistic regression model based on the retained scream-specific principal components achieved a classification accuracy of 0.90, with sensitivity of 0.86 and specificity of 0.94. Discrimination was mainly associated with PC1, which was characterized by amplitude- and energy-related indicators. These findings indicate that, although pen-level localization of all vocalizations remains challenging using a single microphone, acoustic classification of screams may be feasible and could support pen-level behavioural monitoring in farrowing units.
Coutant et al. (Fri,) studied this question.