Nursing behaviour in pigs is vital for piglet survival, as it provides nourishment and essential immunity through colostrum. Disruptions to nursing, which are often caused by health issues such as Postpartum Dysgalactia Syndrome (PDS), can lead to poor growth, compromised health and increased mortality in piglets. Currently, early detection of such disruptions is difficult as monitoring relies on time-consuming and often incomplete manual observations.
This project focuses on developing an audio-based algorithm that can automatically detect and classify nursing events in sows as either successful (indicating milk ejection, characterized by specific call production patterns) or unsuccessful, with no milk ejection. Using synchronised video and audio recordings of sows during the first ten days postpartum — the most critical period for piglet survival — we aim to refine an existing algorithm to accurately detect nursing grunts and build upon it to automatically distinguish between successful and unsuccessful nursing bouts.
This work will lay the scientific foundations for a non-invasive, scalable tool to support the early detection of disease in sows, with the potential to significantly improve the welfare of both piglets and sows, as well as farm efficiency.
Project partner
