You may not think of fish as being noisy, but they can actually be a pretty vocal bunch. A new AI system is able to quickly identify specific fish calls within general reef noise, allowing scientists to better track local populations.
For some time now, stationary underwater acoustic recorders have been used to log the ambient soundtracks of coral reefs as their inhabitants vocalize, eat, dig dens, or otherwise produce noises.
Utilizing signal processing software to analyze the recordings, it is possible to get a general sense of reef activity. If a scientist is seeking the sound of a specific species, however, they typically have to manually search through several months’ worth of acoustic data, painstakingly listening for the call of the sought-after fish in the playback.
The new computer system could change that.
Developed by a team of scientists at the Woods Hole Oceanographic Institution, it utilizes a neural network which was trained on known fish sounds. That network automatically searches through reams of recordings, finding the calls of specific species 25 times faster than is possible for a human listener.
In fact, the system is even capable of matching sounds to species in real time, right as it hears them. This capability could also allow the technology to add previously uncataloged calls to existing databases – if an unknown fish sound were to be picked up by the recorder, an onboard camera could be used to see which fish was making it.
Austin Greene, Woods Hole Oceanographic Institution
With such functionality in mind, the researchers are now integrating the system into an autonomous underwater vehicle known as CUREE. The idea is that the vehicle could cruise over coral reefs, identifying known fish noises and cataloging new ones as it goes.
“For the vast majority of species, we haven’t gotten to the point yet where we can say with certainty that a call came from a particular species of fish,” says the lead scientist, Seth Mccammon. “That’s, at least in my mind, the holy grail we’re looking for. By being able to do fish call detection in real time, we can start to build devices that are able to automatically hear a call and then see what fish are nearby.”
A paper on the study was recently published in The Journal of the Acoustical Society of America.
Source: American Institute of Physics via EurekAlert