University of Victoria Biologists Use AI to Decode Fish Sounds

Biologists at the University of Victoria have made a significant breakthrough in understanding fish communication by using artificial intelligence (AI) to identify sounds made by various fish species. Their research revealed that even closely related fish produce unique and distinctive sounds, allowing for effective differentiation between species.

Using a method known as passive acoustics, the team identified specific sounds from eight fish species native to Vancouver Island. They developed a machine learning model capable of predicting which sounds correspond to which species with an impressive 88 per cent accuracy.

Darienne Lancaster, a PhD student at the university and the lead researcher, highlighted the importance of this discovery. “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,” she stated in a recent news release. “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.”

Unique Sounds and Their Significance

The researchers noted the distinct vocalizations of different species. For instance, the black rockfish emits a long, growling sound reminiscent of a frog’s croak, while the quillback rockfish produces a series of short knocks and grunts. Lancaster remarked on the variety of sounds: “It has been exciting to see how many different species of fish make sounds and the behaviours that go along with these calls.”

Some fish, such as the quillback rockfish, create rapid grunting noises when threatened, suggesting this may act as a defensive mechanism. In contrast, the copper rockfish is observed making knocking sounds as it chases prey along the ocean floor, indicating a different behavioral context.

Innovative Techniques and Global Impact

To identify these fish sounds, Lancaster utilized passive acoustic monitoring, collecting underwater audio and video with a sound localization array. This array was designed by former UVic PhD student and project collaborator Xavier Mouy. The team analyzed the sound characteristics to differentiate species calls effectively.

The AI model developed by the researchers examined a total of 47 different sound features, including duration and frequency. By detecting subtle differences in these features, the model was able to group species calls together efficiently. The techniques established in this research hold potential for scientists worldwide, enabling them to decode other fish calls and enhance understanding of aquatic communication.

The research received funding from the Natural Sciences and Engineering Research Council of Canada and Fisheries and Oceans Canada. This pioneering study not only advances marine biology but also opens up new avenues for ongoing research in aquatic ecosystems.