AUDIE CORNISH, HOST:
Biologists can learn a lot simply by listening. They often record animal sounds. That's easy to do. The hard part is analyzing the sounds. So scientists are working to automate the process, as NPR's Richard Harris reports.
RICHARD HARRIS, BYLINE: Biology professor Mitch Aide uses his ears to learn about the frogs, birds and insects that are all around him. This scientist at the University of Puerto Rico is trying to track species in a world that is under increasing pressure from human activities.
MITCH AIDE: And we would like to have five, 10, 20 years of data of how populations are changing over time.
HARRIS: In recent years, Aide and his colleagues have been putting recorders in the field so they can listen around the clock to the species they are studying. For example, they staked out a marsh where someone recently discovered a new species of frog.
(SOUNDBITE OF FROGS)
AIDE: That is a recording from Sabana Seca and it is a wetland on the north coast of Puerto Rico. This is where we put the original permanent recording station.
HARRIS: It makes 144 recordings a day, every day. It's a great way to keep an ear out for the newly discovered species and half a dozen other amphibians in the area. But there's a catch.
AIDE: So you can save time in the field but in the end, you have to invest a lot of time in the laboratory listening to these recordings. It's torture.
HARRIS: So Aide recruited help from the University of Puerto Rico's computer scientists. As they report in the journal PeerJ, they've developed an automated system to go through all those recordings and pick out the sounds of the species they're studying.
AIDE: The way it works is that the biologist who knows the calls gives the computer examples.
HARRIS: The computer then looks for a match, kind of like the way that Shazam or SoundHound apps can listen to a song playing on the radio and tell you what it is. Aide is encouraging field biologists everywhere to use his open-access system to build up a library of sound-recognition patterns that researchers can share.
AIDE: Right now, anybody that has Internet connection can listen and view over a million recordings that we've collected from the different projects.
HARRIS: And what's especially appealing about this is he's gradually building a permanent repository of field recordings.
AIDE: One of the things I really like about this technology is that each recording is like a museum specimen. Anybody in 10, 15, 20 years can go back with new technology, with new ideas, and listen and analyze those recordings again.
HARRIS: Aide's project is just one example of computerized sound recognition in biology. Cornell University has a big project focused on the world's oceans.
(SOUNDBITE OF FISH)
HARRIS: Here's an example. They recorded a fish chatting off the coast of North Carolina.
AARON RICE: We have networks of recording devices around the world listening for fishes and whales as far north as the Arctic Circle, down to the tropics in the Atlantic and Pacific oceans.
HARRIS: Aaron Rice at Cornell says that means they are also swamped with recordings.
RICE: In our lab group, we have years of data that no mere mortal could ever listen to in a single lifetime.
HARRIS: So he, too, is working to automate this process by developing computer programs to sift through all that sound to identify animals and track them as they move throughout the world's oceans. Rice says the same idea is also taking hold across the parched lands of Australia.
RICE: They have a string of recorders going from north to south across Australia and their goal there is to identify the spread of invasive bird species.
HARRIS: That project also illustrates one of the challenges of these computerized systems. It turns out, one of the most aggressive invasive species is the European starling. And, as this recording from Cornell's Macaulay Library shows, starlings like to mimic other species.
(SOUNDBITE OF BIRDS)
HARRIS: It's not easy to teach a computer to recognize those ambiguous sounds. And the task at hand is much larger. The scientists would ultimately like computer programs to identify every animal sound that appears in a recording, so they can track the fate of the world's species. Richard Harris, NPR News.
NPR transcripts are created on a rush deadline by Verb8tm, Inc., an NPR contractor, and produced using a proprietary transcription process developed with NPR. This text may not be in its final form and may be updated or revised in the future. Accuracy and availability may vary. The authoritative record of NPR’s programming is the audio record.