Artificial intelligence that detects and quantifies distress calls

Artificial intelligence that could improve the welfare of farmed chickens by eavesdropping on their squawks could be available within five years, researchers say.

New research suggests that the technology that detects and quantifies distress calls made by chickens housed in huge indoor sheds distinguishes distress calls from other barn noises with 97% accuracy. A similar approach could be used to drive up welfare standards in other farmed animals.

Each year, around 25 billion chickens are raised, many of which are kept in huge sheds, each housing thousands of birds. One way to assess the welfare of such creatures is to listen to the sounds they make.

Chickens are very vocal, but the distress call tends to be louder than the others, and is what we would describe as a pure tonal call, said Alan McElligott, an associate professor of animal behaviour and welfare at the City University of Hong Kong. It is not too hard to pick them out, even to the untrained ear. In theory, farmers could use chickens calls to gauge their level of distress and enrich their housing where necessary. Human observers are not practical in commercial flocks containing thousands or tens of thousands of chickens. Their presence could stress the flock, but with so many birds, objectively quantifying the number of distress calls is impossible, McElligott said.

His team has developed a deep learning tool to automatically identify chicken distress calls from recordings of intensively farmed chickens. The tool was trained using recordings that had already been classified manually by human experts to determine what type of sound they represented.

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

McElligott said the technology could be commercially deployed within five years, but our goal is not to count distress calls, but to create conditions in which the chickens can live and have a reduced amount of distress.

Before that happens, the team will need to make sure that the recording equipment works in different types of chicken sheds, as well as testing it in farms with higher or lower welfare standards to confirm that the readings correlate.

It's relatively easy to persuade farmers to adopt the technology. Previous research done by McElligott found that distress calls made by young chicks could predict the amount of weight gained and the number of deaths in the whole flock during its lifetime.

It is sometimes hard to convince farmers that they have to deal with the problem of producing animals for a set price for supermarkets and everyone else to adopt technology to improve their welfare, McElligott said. This is a way of automating the process because we already shown that distress calls are a good indicator of mortality and growth rates. He said similar technology could be developed to monitor other farmed animals, particularly pigs or turkeys, which are often housed indoors and are highly vocal.

The research was welcomed by the RSPCA. This technology can be incredibly useful in monitoring and improving the welfare of farm animals, but we wouldn't want to see this replace physical inspections or reduce stock keeper-bird contact, as this could lead to a loss of stockmanship skills or birds that are more difficult to handle, a spokesman said.

The vocalisation of distress is just one welfare indicator, but there are a number of other physical factors such as lameness and leg burns that farmers should look out for.