An international team of researchers can use artificial intelligence to interpret the grunting sounds of pigs. This could lead to a revolution in animal husbandry. […]
Grunting, sniffing, squealing – the sounds caused by pigs give indications of the emotional state of the animals. With an AI translator, scientists have now managed to find out more details. In the best case, the experiment could lead to automated monitoring systems that can be used to monitor the well-being of farm animals. The term “animal welfare” would then have a completely different meaning.
“We can decode the emotions of pigs,” says a message from the University of Copenhagen. “We collected thousands of acoustic recordings during the life of pigs, from birth to death. An international team of researchers has translated the grunt into the emotions that these animals express.“
In addition to the University of Copenhagen, ETH Zurich and the French National Research Institute for Agriculture, Food and the Environment (INRAE) as well as other researchers from Germany, Norway and the Czech Republic are involved. The first research results were published in the “Scientific Reports” on Nature.com published.
The researchers evaluated exactly 7414 sound recordings of 411 pigs – both in a laboratory environment and in commercially set up stables. To do this, they have developed a machine learning algorithm that decodes whether a pig shows a positive emotion (“happy”, “excited”) or a negative emotion (“scared”, “stressed”). There are many more gradations between these extremes.
Subsequently, it was possible to separate the positive from the negative sounds, with high-frequency sounds expressing rather negative emotions. Surprisingly, a deep grunt does not always mean satisfaction, in certain variations it can also signal anxiety and stress. Of particular interest was the wide range of tones between these two extremes. The analysis of the noise patterns revealed detailed information about what happens in the animals in a wide variety of life situations – from suckling at the mother’s teat to transport to the slaughterhouse.
“In positive situations, the grunt sounds are much shorter, the amplitudes on the noise scale are smaller, ” says Elodie Briefer, assistant professor of biology at the University of Copenhagen and lead author of the research. The height, volume and spacing of the grunt give information about the psychological state. “By training our algorithm to recognize the sounds, we can correctly assign sounds and emotions with 92 percent certainty,” says Briefer.
The next step is to develop the algorithm into an app so that farmers can improve the welfare of their animals. The physical monitoring of animal populations has already been successfully tested in many experiments, with the control of psychological well-being, the researchers are entering new territory.