The technology can identify a stressed chicken's call from the noise of the coop with 97 percent accuracy.

"Chickens are very vocal animals" : Researchers have found that chickens make louder than usual noises when they are stressed.

According to the researchers, artificial intelligence technology can capture different calls of chickens to analyze their welfare status. The technology is expected to be ready for use within five years.

The technique detects and quantifies the calls of chickens raised in a house under stress, and can distinguish them from other noises with 97 percent accuracy. The research suggests that the technology could also be used to improve the welfare of other farm animals.

Globally, about 25 billion chickens are raised each year, and the vast majority of them are raised in houses, with thousands of chickens in each house. So one way to assess the welfare of chickens is to listen to their calls.

Alan McElligott, associate professor of animal behaviour and welfare at the City University of Hong Kong, said: "Chickens are very vocal, and when they are stressed, they will call more loudly, which is what we call a tonal call, which is not very difficult to recognise even for untrained people."

Alan McElligott said: "In theory, farmers could use the sounds of chickens to gauge their stress levels and improve their conditions if necessary. However, it is not practical to rely on human observation to manage hundreds of thousands of commercially farmed chickens. "In some ways, the presence of humans may cause more stress to the chickens, and it's impossible to detect stress calls effectively in such a large breeding population."

So his team has built a deep learning tool that can automatically identify the calls of chickens from intensively farmed flocks. The tool is trained to use recordings that have been categorized by experts so that it is clear what the different sounds represent.

According to an evaluation published in the Journal of the Royal Society Interface, the algorithm correctly identified 97 percent of the stress calls.

"Our goal is not to count how many stressful calls there are, but to put the chickens in a less stressful environment, and this technology could be available within five years," says McElligott.

Before starting to work on the algorithm, the team had to make sure that the recording device could be run in different types of coop, and that it had to be tested in high-welfare or low-welfare coops to confirm the welfare status of different calls.

Persuading farmers to adopt the technique may be relatively easy; previous research by Alan McElligott has found that stress calls from chicks can predict weight gain and mortality over the life cycle of a flock.

McElligott added: "It is sometimes difficult to convince farmers who have to produce animal products for supermarkets or consumers at a fixed price to use technology to improve animal welfare, but we have shown that stress calls are a good predictor of mortality and growth rates, and can also automate farming processes." Similar techniques could be developed to monitor other farm animals, especially animals like pigs or turkeys that are often kept indoors and tend to make noises."

A spokesman for the RSPCA said: "We very much recognise this research and studies like this are very useful in detecting and improving the welfare of farm animals. But we do not want to see this technology replace on-site inspections or reduce contact between the animal and the breeder, as this can lead to a loss of breeder skills or more difficult handling of the animal. In addition, stress calls are only one indicator of animal welfare, but there are other physiological indicators such as lameness and leg burns that should be looked at."