With 42 different species of dolphin using hundreds of different sounds to communicate, you’d think researchers would have their work cut out enough discerning one voice from another. Add to the matter that many dolphins adopt their own regional ‘accents’ and it’s little wonder scientists have turned to technology to cut through the chatter.

Experts at the University of St Andrews Scottish Oceans Institute are harnessing machine learning to break new ground in dolphin communication and develop a tool that can identify different species based on their calls. 

It has been created by the University’s Dr Julie Oswald, and is a tool so advanced it can even provide different modules for these regional ‘accents’ that dolphins have been found to adopt. Not only that, it can categorise dolphin calls by their species with different versions linked to different geographical regions.

It’s benefit is already being felt. The underwater world is a noisy environment, with dolphins being responsible for a great deal of that chatter. There are around 42 species of dolphin using hundreds of different sounds to communicate. From a young age, dolphins learn to mimic the sounds they hear while it’s recorded that some even develop a unique “signature whistle” that they use to announce their identity throughout their lives.

Because dolphins are so acoustically active, scientists can study them using hydrophones – a type of underwater microphone that captures sounds as dolphins pass by. Analysing these sounds, however, has historically been far from easy.

“We can use visual observations and hydrophones to capture audio, but we can’t easily match those sounds to specific dolphin species,” said Dr Oswald. 

This is where the new dolphin translator technology comes into play. It’s known within certain circles as the Real-time Odontocete Call Classification Algorithm (or ROCCA for brevity) and it overcomes the challenges faced by using machine learning to uncover the subtle acoustic differences in dolphin species’ chatter. 

From here, it can match dolphin vocalisations to species with regionally specific settings to account for their varying accents. This allows the tool to identify the species present in a particular area.

Like humans, dolphins develop region-specific accents, meaning that a bottlenose dolphin in the North Sea may produce different sounds compared to one in the Pacific. ROCCA accounts for these differences, enabling conservationists to more accurately identify dolphin species and assess the impact of human activities, such as fishing and sonar disturbances.

So successful has the tool proven, Dr Oswald is now expanding ROCCA’s applications to other species and ecosystems in regions like West Africa, Macaronesia, and the Mediterranean while working with other scientists and conservationists to train them in acoustic monitoring using the tool.

In early December, Dr Oswald’s ROCCA was named one of two winners of the Earth Rangers 2024 Conservation Technology Awards, an accolade that recognised innovations and technology-driven solutions that make measurable conservation impacts. 

Now in their fourth year, the Awards have provided some $120,000 in funding to conservation organisations across the globe.

Jes Lefcourt, senior director of conservation technology at Ai2 and director of the EarthRanger programme, said: “The Scottish Oceans Institute has been improving the understanding of cetacean communication for 20 years. Their recent innovations using AI have improved the ability to track and monitor populations of these past species and also highlight the individuality of the animals.”

Upon receiving the award, Dr Oswald said she was ‘delighted to be recognised in this way’ and will use the £15,000 fund from the award to ‘help create new classifiers, build data collection capacity, and train more people to use ROCCA.’

Click here for more from the Oceanographic Newsroom.