Algorithm helps translate grunts and squeaks

Pigs express their emotions by grunting and squeaking. Thanks to an algorithm, those sounds are now easier to interpret.

Pig farming regularly makes the news with abuses, such as barn fires that kill thousands of pigs. Nevertheless, efforts are also being made to improve the welfare of the animals that we eat en masse. An example of this is the so-called family barn, where pigs stay with their relatives, with more space and a stimulating environment without concrete floors.

Pig farmers can also hear how their animals are doing. Squeaking and screeching pigs are in pain or in fear, while various grunts indicate contentment, curiosity or heat, meaning a sow wants to be mated.

ALSO READ
Giant Neural Networks

In some pig houses there are therefore microphones that can detect an anxious mood among the animals. In addition, the microphones are intended to pick up coughing and coughing from sick pigs.

Elodie Briefer, animal welfare researcher at the University of Copenhagen, decided to find out which pig sound means what. She did this in collaboration with animal welfare researchers from Switzerland, France, Germany, Norway and the Czech Republic. Over a period of fifteen years, they collected 7,414 pig sounds from 411 pigs, made in research institutions, as well as commercial pig farms, at all stages from birth to slaughter.

Drinking with the mother

In the magazine Scientific Reports they analyze the sounds acoustically and compare that data with the conditions under which they were grunted. Positive circumstances were, for example, the moment when piglets drank from the mother, reunion with related pigs or playing with toys in a so-called enriched environment. Negative situations were fights between piglets, jostling in which animals are oppressed, being separated from family and – the final destination of almost every pig’s life – the last moments in the slaughterhouse.

The behavior of the animals themselves also indicates how they feel. It was already known that pigs stand for a long time when they are afraid, while they turn their ears forward when they feel good. To evoke more nuanced pig feelings, such as curiosity and playfulness, the researchers created spaces with toys and food.

Automatic classification

As expected, beeping and screaming turned out to indicate negative situations. However, a low grunt does not always mean that pigs are feeling well. For example, grunting often or for a long time indicates fear. “In positive situations, the grunts are much shorter and do not vary much in volume,” Briefer, the study’s lead author, said in a press release.

The researchers trained a self-learning algorithm to recognize the pig sounds and classify them as positive or negative expressions of feeling. That worked very well. Despite disturbing noises from, for example, fans or reverberation in the pigsties, the algorithm was able to correctly determine the difference between positive and negative in 92 percent of the cases. That’s just as good or better than current systems, which only detect fear and illness.

Briefer: ‘Now we need someone who wants to further develop the algorithm into an app that pig farmers can use to improve animal welfare.’

ttn-15