UVic researchers use AI to decipher fish sounds

University of Victoria (UVic) biologists have discovered that even closely related fish species make unique and distinctive sounds and determined that it’s possible to differentiate between the sounds of different species. The discovery opens the door to identifying fish based on sound alone.  

Using passive acoustics, the researchers identified unique sounds for eight different Vancouver Island fish species in their natural habitats. They then developed a machine learning model that can predict which sounds belonged to which species with up to 88 per cent accuracy. This could have positive implications for marine conservation efforts and allow scientists to monitor specific fish species using acoustics, says Darienne Lancaster, a PhD candidate in biology who led the project. 

We knew previously that many fish were making sounds in the wild, but we didn’t know which sounds belonged to which species, or if it was possible to tell these sounds apart. Now, just as we use bird song to identify specific bird species in the wild, we can also listen to fish sounds to identify specific fish species.” 

—Darienne Lancaster, UVic PhD student and lead researcher

The research, published in the Journal of Fish Biology, is part of the larger fish sounds project run out of the Juanes Lab at UVic. 

Fish grunts, knocks and growls  

While researchers have been identifying fish sounds for years, these sounds were typically recorded in a laboratory setting, rather than in the wild and whether different species made unique sounds had never been tested.  

Lancaster identified unique sounds for eight different species of fish commonly found on the coast of British Columbia: the Black Rockfish, Quillback Rockfish, Copper Rockfish, Lingcod, Canary Rockfish, Vermillion Rockfish, Kelp Greenling and Pile Perch. This was the first time, in the lab or the wild, that sounds had been identified for the Canary and Vermillion Rockfish. 

“It has been exciting to see how many different species of fish make sounds and the behaviours that go along with these calls,” says Lancaster. “Some fish, like the Quillback Rockfish, make rapid grunting sounds when they’re being chased by other fish, so it’s likely a defensive mechanism.  Other times, fish, like Copper Rockfish, will repeatedly make knocking sounds as they chase prey along the ocean floor.” 

The Black Rockfish make a long, growling sound similar to a frog croak and the Quillback Rockfish make a series of short knocks and grunts.  

The role of passive acoustics

Two scuba divers setting up the sound localization array used to capture underwater audio and visual. Credit: Shane Gross.

Lancaster used a technique called passive acoustic monitoring to identify the fish sounds. She collected underwater audio and video using a sound localization array designed by former UVic PhD student and project collaborator, Xavier Mouy, and then used sound characteristics to identify differences in species calls.  

Her machine learning model used a set of 47 different sound features, such as duration and frequency, to detect small differences in each species’ sounds that can be used to tell them apart. The model used these small differences in sound features to group species calls together. 

“The ability of passive acoustics to identify specific fish by sound could be an important new tool for conservationists and fisheries managers,” says Francis Juanes, UVic biology professor and principal investigator on the project. “Passive acoustics could allow us to estimate population size, monitor activity, and assess the overall health of a fish population in a way that is minimally invasive to vulnerable marine animals.” 

The techniques that Lancaster developed can be adapted by scientists all over the world to decipher other fish calls.  

The Fish Sounds project

The Juanes Lab fish sounds project includes: FishSounds.net, an online database of global peer-reviewed fish sounds research; The Codcast, a podcast about the discovery of the Arctic cod grunt; the development of a sound localization array used to simultaneously capture sound and video; and Fish Sounds Educate, a free educational program that aims to use the topic of bioacoustics to encourage future conservation leaders and enhance ocean literacy across Canada.  

The research was funded by the Natural Sciences and Engineering Research Council of Canada and Fisheries and Oceans Canada and supports the United Nations Sustainable Development Goals 14 (life below water). Learn more about SDGs at UVic.