University of Victoria Biologists Harness AI to Identify Fish Sounds

Researchers at the University of Victoria have made a significant breakthrough in understanding fish communication. They discovered that even closely related fish species produce distinct sounds, enabling them to differentiate between these species acoustically. By employing passive acoustics, biologists identified unique sounds for eight fish species native to Vancouver Island.

Utilizing a machine learning model, the team achieved an impressive 88 percent accuracy in predicting which sounds correspond to which species. Lead researcher and PhD student Darienne Lancaster stated, “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.” This advancement parallels how ornithologists use bird songs to identify different bird species in their natural habitats.

The study detailed various fish sounds, including the long, growling noise of the black rockfish, which resembles a frog croak, and the series of short knocks and grunts made by the quillback rockfish. Lancaster noted, “It has been exciting to see how many different species of fish make sounds and the behaviours that go along with these calls.” For example, the quillback rockfish emits rapid grunting noises when threatened, suggesting it employs these sounds as a defensive mechanism. Conversely, the copper rockfish produces knocking sounds while chasing prey along the ocean floor.

Innovative Techniques in Marine Biology

Lancaster employed passive acoustic monitoring to gather underwater audio and video. This technique utilized a sound localization array designed by Xavier Mouy, a former UVic PhD student and collaborator on the project. By analyzing sound characteristics, the researchers differentiated species calls effectively.

The machine learning model developed by Lancaster used a comprehensive set of 47 sound features, including duration and frequency, to identify subtle differences between the calls of various fish species. These small distinctions allowed the model to group calls according to their specific species. The methods pioneered in this study have the potential to be adapted globally, aiding scientists in deciphering other fish calls.

The research received funding from the Natural Sciences and Engineering Research Council of Canada and Fisheries and Oceans Canada, highlighting the importance of government support in advancing marine biology research.

As scientists continue to explore the underwater soundscape, this innovative approach may offer new insights into fish behaviour and communication, ultimately contributing to more effective conservation strategies. The ability to identify fish species through their sounds could transform how biologists study marine life, paving the way for future discoveries in aquatic ecosystems.