Sows produce characteristic vocalizations during estrus, including short high-pitched calls and longer rhythmic grunts. Recognizing these sounds helps farmers determine the appropriate breeding time. In commercial farms, background noise and the relatively short duration of estrus can make these vocal signals difficult to detect. This study presents a sound-based monitoring approach that distinguishes estrus calls from routine farm noise. Compared with camera-based systems, acoustic monitoring is less influenced by lighting conditions, animal posture, or visual obstruction within the pen. The system can operate on standard farm computers without requiring specialized hardware. To improve training reliability, computer-generated estrus sounds were used to supplement real recordings. This strategy reduces dependence on extensive manual data collection while strengthening model robustness. The approach provides a non-invasive method for continuous monitoring of sow reproductive status and maintains stable performance in noisy farm environments. It may contribute to improved breeding management and more efficient farm operation.

Read the full article in the Journal of Animal Science.