An in-depth learning model can extract chicken emergency calls from recordings made on commercial farms and can be used to improve chicken welfare
June 29, 2022
An AI is trained to identify and count the distress calls of chickens. Farmers could use the tool to improve conditions for chickens raised on busy commercial farms.
As of 2020, there were more than 33 billion chickens around the world, according to the Food and Agriculture Organization of the UN. Many of these animals live in poor conditions, crammed together with little ability to move or do things that chickens love to do. “Despite the fundamental concerns about not being hungry or thirsty, there are still serious concerns about how they are produced,” says Alan McElligotto at the City University of Hong Kong.
According to previous research by McElligott, the frequency and volume of a chicken’s distress call — a sharp, short “beep” — can predict the animal’s health and growth rate. But he says these calls are hard to spot when there are thousands of chickens howling at each other — in some barns, it could be 25,000 or more. “They’re called sheds, but they’re more like airline hangers,” says McElligott.
McElligott and his colleagues listened to recordings made on large broiler farms in southwestern China and mentioned emergency calls for chickens, distinguishing them from farm noises and other chicken sounds like chirps of pleasure and trills that indicate fear. With this labeled data, they trained several algorithms to identify emergency calls of the background noise and measure their frequency and volume. When tested on other tagged shots from the same farm, the best algorithm accurately detected emergency calls about 85 percent of the time.
The tool has not yet been used on a working chicken farm, and McElligott says there is more work to be done to understand the link between emergency calls and a chicken’s well-being. But he says the next steps are somewhat obvious: “We should give them conditions where, well, maybe they would produce fewer emergency calls.” This may mean giving chickens more space or other enrichment, such as straw bales to scratch and climb on.
Elodie Floriane Mandel-Briefer at the University of Copenhagen in Denmark has developed similar tools to assess the emotions of pigs based on sounds and facial expressions. She says the research in chickens adds to growing evidence that: animal emotions can be measured and controlled using machine learning. “Because emotions in animals are an important part of their well-being, their assessment is crucial,” she says.
Reference magazine: Journal of the Royal Society Interface† DOI: 10.1098/rsif.2021.0921
More on these topics: