New tech can tell a squawk from a cluck
Stockmanship skills could be eroded if technologies were relied on too much. Jim Webster
AgResearch animal welfare researcher
Researchers from Hong Kong and the United Kingdom have developed a deep learning tool to identify the distress calls of farmed chickens.
Using recordings from an intensive chicken farm, the team developed an algorithm that identified 97% of distress calls.
As chicken were often housed in groups of thousands, automated methods would be helpful to monitor distress calls, the researchers said.
They said more than 25 billion chickens were farmed worldwide each year, and monitoring chicken noises for distress could be an easy way to ensure chickens received veterinary care when they needed it.
But AgResearch animal welfare researcher Jim Webster said stockmanship skills could be eroded if technologies to identify animal welfare issues were relied on too much and created separation between human and animals. ‘‘People still need to go in and carefully walk the sheds,’’ Webster said.
This technology acted on a flock level and would not be able to pick out a single bird that might be lame, for example.
But chickens were very vocal about their level of comfort and such technology could be employed to detect flock-level stress in case feeding belts, water supply or ventilation systems failed, Webster said.
Technologies could make flocks appear to be single units or something to be managed like a machine, with food going in on one side and eggs or meat chickens coming out the other side.
This could detract from the recognition of sentient animals. Animal welfare laws required recognition of sentience which was at the level of individual animals, Webster said.
However the poultry industry used a lot of technologies to monitor feeding, breeding and living conditions and this technology could become another tool to help with animal welfare, Webster said.
Research showed that a larger dataset needed to be built that included more types of vocalisations from different breeds and production environments in the future.
The model could be used with other detection methods to achieve additional functions, such as identifying the vocal source location.
The authors said chickens’ early-life welfare constraints often predicted later-life welfare concerns. The output of distress vocalisations in commercial flocks was linked to growth rates and mortality levels.
The technology still had challenges to overcome.
In real-world identification, many other specific chicken sounds, such as alarm and gakel calls, might also be regarded as potential welfare indicators, and distress calls might be inconsistent between different breeds and welfare scale systems.