Vocalization Study Provides Insight into the Well-Being of Birds in Growout Houses
The Growout Monitoring System collects and then analyzes bird vocalizations to determine if the birds are behaving in an unusual manner due to environmental conditions, disease, or other stressors. To view a video of the project, scan the QR code above.
Sensors capable of reliably monitoring the well-being of birds reared in confined housing currently do not exist. Two years ago, engineers with the Georgia Tech Research Institute (GTRI) and poultry scientists at the University of Georgia (UGA) set out to investigate whether the birds themselves could be the answer. Anecdotal evidence and previous research at the University of Connecticut indicated that this might be a possibility. So, to help the poultry industry assess growout house conditions and thus flock health and performance, the team built an experimental monitoring system. The Growout Monitoring System collects and then analyzes bird vocalizations to determine if the birds are behaving in an unusual manner due to environmental conditions, disease, or other stressors.
“The well-being of animals is of significant concern to the poultry industry. Currently, there are no quantitative techniques for measuring the state of animals in confined environments. The objective of our research is to determine vocalization features that can be extracted and used as a measure of animal well-being,” explains Dr. Wayne Daley, associate division chief of GTRI’s Food Processing Technology Division and project director.
Researchers start by isolating bird vocalizations that are of interest. They then select features that might help
with identifying those vocalizations and differentiate them from background noises. Such features could be the
sound levels of the audio at various frequencies. For example, birdcalls, sneezes, and coughs are made up of different frequencies or tones, so those are characteristics of the sound that can be used to identify them. Thus, a sneeze would consist of higher frequencies, while a cough would have lower frequencies.
Next, researchers use machine learning techniques to evaluate the important features along with corresponding algorithms to identify the vocalizations. The machine learning system is taught to differentiate the isolated bird vocalization sounds from the background noise and other sounds by processing different examples of a particular bird vocalization as well as examples of sounds that are not the vocalization. The anticipated result is that features can be found that represent unique “syrinxprints” (the syrinx is the equivalent in a bird of the voice box or larynx in a human) of the vocalizations as well as algorithms to identify the vocalizations matching those features.
The team can then use those features and algorithms to automatically identify vocalizations that are made by the birds when they are sick or under stress due to environmental conditions within the growout house.
Recent studies conducted at UGA’s Poultry Science research farm and its Poultry Diagnostic and Research Center (PDRC) supported the use of this approach. The team studied environmental effects such as temperature, ammonia, and crowding
in the farm’s growout house. Results showed that features extracted from bird vocalizations correlated with higher ambient room temperatures and the presence of ammonia. However, the correlations for crowding were not as conclusive. Experiments conducted at the PDRC explored the effects of two broiler diseases, Infectious Bronchitis and Laryngotracheitis (LT). In both experiments, algorithms identified vocalizations that correlate with the progress of both diseases.
Researchers are now attempting to replicate these results and investigate applicability in a commercial setting. They are also seeking to identify other features of interest that correlate with animal and growout house conditions.
“Our ultimate goal is to develop techniques that allow us to extract features that provide a more quantitative measure of animal well-being that could lead to more effective management practices. For example, if we are able to detect diseases early in the growout cycle, it provides managers with more options to keep the flock healthy,” says Daley.
The team hopes to define a path for future commercial application of the research results.