AI tracks individual ovenbirds by their signature songs
A new study uses artificial intelligence to identify individual ovenbirds from their songs with up to 96 percent accuracy.
Biologists are using artificial intelligence to track the health of ovenbird populations by identifying individual birds from their signature songs alone. Research published this month in the journal Nature Methods used AI to analyze recordings from Alberta and Pennsylvania, linking songs to specific birds with up to 96 percent accuracy. The ovenbird's stable, single-song repertoire made it an ideal species for the study. The method offers a less labour-intensive alternative to traditional leg banding for monitoring these ground-dwelling warblers.
"They are really pretty loyal to singing one very specific song," said lead researcher Sam Lapp, a postdoctoral researcher at the University of Colorado. He noted that while many birds develop a vocal signature over their lives, the ovenbird has the unique property of singing only one, very stable version of its song throughout its life. The team, led by Lapp during his time at the University of Pittsburgh, processed 9,072 hours of audio from nearly 130 sites in Pennsylvania over four years to track 405 individual birds. The AI model was trained to extract features that made songs unique and cluster similar ones together, operating on the assumption that each bird returns to the same forest patch each spring. The researchers then tested the model against another set of recordings collected in Alberta's mixed wood and boreal forests during the 2016 spring breeding season. Lapp said the idea for the research came from a collaboration with Alberta biologists who were trying to pinpoint bird locations using a grid of microphones.
Human ears cannot distinguish individual ovenbirds in real time because their vocal nuances happen too quickly. However, on spectrographs, which are visual representations of sound, each bird's distinctive song becomes clear. "I started thinking, if my eyes can see these patterns, can we teach a machine to be able to recognize these patterns and do an automated version of individual recognition?" Lapp said. He described the discovery of this level of detail in bird song as surprising and wondrous.
READ MORE: Some workers replaced by AI are being asked back by their employers
READ MORE: Canadian government websites go offline to patch critical PHP vulnerability
Erin Bayne, a University of Alberta biology professor and co-author of the study, has spent years monitoring ovenbirds. During his PhD, he spent tens of thousands of hours tracking banded birds and learned to tell individuals apart by their unique vocal variations. "What that guy is saying over there is a little different from that guy over there," he said. "I knew who was who." Bayne noted the birds have polite singing manners, with one waiting for another to finish before singing instantaneously after.
Bayne said the study shows avian communication is far more complex than most people assume, with conversations happening even if we do not understand the language. He compared it to humans having distinct speech patterns, suggesting that as researchers look more closely, they will find it is common for individual animals to have detectable differences. The researchers have put the program code on open-source sites to help other biologists track the survival of other songbird populations or even other species. The AI system has some limitations. It cannot accurately track the varied calls of female ovenbirds or the simple vocalizations made by juveniles. The current program would also fall short for monitoring species like magpies and mockingbirds that mimic new sounds or sing dozens of different songs. It would work best for the roughly 30 percent of songbird species that sing one, unchanging song throughout their lives, as well as certain species of bats and owls.
Richard Hedley, a statistical ecologist with the Alberta Biodiversity Monitoring Institute who was not involved in the research, said the study has incredible potential. He noted that while AI acoustic monitoring has been used for years to track birds at the species level, showing it can distinguish between individual birds is an impressive feat. Hedley said there is a wealth of existing acoustic recordings that could be harnessed in new ways. "It can take many hours just to catch a single bird. So the idea that we can just do this by putting a microphone on the bird's territory is really promising," he said.
With files from CBC News