Broiler chicken vocalization analysis during a medium-scale heat stress experiment

The poultry industry in Flanders, Belgium, is characterized by highly efficient production mechanisms. To improve awareness for animal health and welfare, the use of neural network-based broiler vocalization detectors has been recently proposed in scientific literature. However, these tools have not yet been tested when used continuously in large-scale deployments. This paper describes the deployment of the broiler vocalization detector to analyze the sonic environment during four experimental rounds, each containing 560 broilers, in a medium scale housing at ILVO, Belgium. Half the broilers were subjected to periods of heat stress (ambient temperature increased to 32°C for 6h/day for 10 days during the last weeks of their lives) and the effect on the diurnal broiler vocalization pattern and occurrence of warbles, pleasure notes, short beeps and distress calls is discussed. While these four different types of vocalizations are common vocalizations in young birds, new types of vocalizations are emerging at an older age. A clustering has been performed to further analyze these new types of vocalizations occurring.