A groundbreaking research project compiles a comprehensive database of pig sounds and utilizes AI to accurately interpret them, heralding advancements in farm animal wellbeing.
In an industry-first, animal behavior research is leveraging an AI environment to decode the complex language of pigs, offering the potential to significantly advance farm animal welfare. This pioneering project gathered a plethora of pig noises throughout different stages of their life, aiding farmers in gaining a deeper understanding of their livestock’s emotional well-being.
AI Environment Decodes Grunts
Utilizing artificial intelligence, researchers were able to accurately determine the emotional nuances of pig sounds such as grunts, squeals, and oinks. The AI environment aided in not only identifying positive or negative emotional states, but also pinpointing the precise scenarios in which these sounds occurred, including mother-offspring interactions, waiting periods, surprise reactions, and huddling in groups.
Elodie Briefer, an animal behaviorist at the University of Copenhagen, spearheaded this study which centred around pigs due to their remarkable vocal range. A staggering 7,414 pig sounds from 411 different animals, recorded since 2013, were analyzed, including those from nursing piglets, weaned piglets, and adult pigs in isolation. The team expanded the database by adding new recordings, ultimately compiling the largest audio catalogue of pig sounds ever, essentially covering a pig’s entire lifespan.
Deciphering the pig language was the next hurdle. To this end, machine-learning methodologies were employed to categorize each grunt as positive or negative and establish its contextual relevance. The database covered 19 unique contexts featuring distinctive pig grunts, such as those made during running, fighting, and waiting.
Contrary to conventional wisdom suggesting high-pitched sounds are negative and low-pitched ones are positive, the AI environment identified several anomalies. Some high-pitched, short duration calls that maintained a consistent volume were, in fact, interpreted as positive.
The most accurate interpretations were achieved by a neural network which analyzed the call frequency in relation to its duration. The AI system achieved a commendable success rate, accurately determining the emotional content and the context of a call 91% and 81% of the time, respectively.
The potential benefits of these findings could extend globally, impacting millions of pigs in farming. Briefer’s team plans to create a software application leveraging this comprehensive database to oversee commercial pig herds. This application would alert farmers when their animals display high degrees of negative emotions, fostering more empathetic and responsive animal care.
The extensive nature of the research earned praise from animal behaviorist Alan McElligott from City University of Hong Kong. He emphasized the importance of understanding an animal’s complete lifecycle from piglet to slaughter as it assists in differentiating between vocalizations in diverse emotional states.
McElligott expressed optimism that this AI Environment tool could prove instrumental in enhancing pig welfare, a pressing global issue considering the sheer numbers involved in commercial farming.
