Machine listening can help us understand animal behavior

- Machine listening gives us the ability to listen to our biodiversity. It allows for large-scale mapping of acoustically active species over long periods of time. This provides insights into how populations change over time and can also help map animal language and behavior. Many species are currently undergoing a mass extinction and we need tools that help us understand the right way forward,” says John Martinsson, researcher at RISE. 

One source of information about species in different ecosystems is sound. Sound analysis can be done with the help of artificial intelligence, something that a research group at RISE specializes in.

- Sound analysis works on sounds in nature in general but is particularly useful for learning more about acoustically active animals, such as birds, insects, bats and wolves. Sound can be used to distinguish species, numbers, location and their health in a specific area,” explains Olof Mogren, research leader at the RISE, Center for Applied AI. 

Collaboration between researchers and ecologists

How species feel, move and reproduce are often key questions when it comes to major social decisions.  For example, new wind farms, roads or other changes in nature. That’s why AI research at RISE is important.

- Identifying a problem together is the first step in the collaboration between ecologists and us AI researchers. The dialog opens our eyes from both perspectives! We at RISE contribute to how ecologists can work more efficiently with the data they collect, and ecologists, in turn, can respond to what they want to find in the sound, such as indications of a stressed animal,” Olof explains. 

Areas of use 

Audio recordings can be made at long distances and are therefore powerful for data collection, compared to visual recordings where the species must be close to the camera. By combining audio recording and machine learning, researchers can gain deeper insights into wildlife and contribute to environmental work. For example, sound analysis can be used for: 

• Species and population identification

Audio recordings can be analyzed to identify species and monitor populations. For example, automated algorithms can distinguish between different bird species based on unique song patterns. 

• Behavioral studies

Sound analysis can help to understand species movements and behaviors. For example, by studying variations in bird song, researchers can gain insights into mating rituals, territorial behavior, and social interactions. 

• Habitat and ecosystems

Sounds can be used as markers to draw conclusions about entire ecosystems. By studying the acoustic diversity of an area, scientists can assess the health and status of biodiversity in the area. 

• Movement patterns

Automated recording stations can be placed along set routes to collect data on when and where different species pass. 

• Climate change

By analyzing changes in sound over time, scientists can draw conclusions about the effects of climate change, such as changes in habitat and species abundance.

• Archiving natural soundscapes

Sound recordings allow us to archive the history of the Earth in sound, making it structured and searchable, which is essential to draw insights from the archived sound. 

• Mapping the language of animals 

The first tentative conversations with whales have been made possible partly by the technologies we are researching.

• Human impact on animals

It is possible to map how human noise interacts with where animals choose to live and how it affects their well-being. 

RISE techniques for sound analysis

There are several tools and methods used in modern sound analysis of species.  Researchers use a combination of field microphones, sensors, spectrograms and machine learning algorithms. With specialized field microphones or sensor networks, large amounts of data can be collected continuously over geographical areas. Sound analysis is then performed on the spectrograms that visually represent the sound with frequency over time.  It can be used to identify specific song patterns in acoustically active species. Machine learning algorithms can then be used to automate the identification of species based on this audio data. 

- The AI can determine if a specific species is present in the area, but also provide a more fine-grained analysis that assesses which individual it is and what it should do,” explains Olof.

In conclusion, modern sound analysis can be powerful tools for understanding species and their ecosystems. By combining sound recording and machine learning, researchers can provide deeper insights into bird and other species and contribute to their conservation and informed decision-making.

Related articles

AI ger oss möjligheten att lyssna till vår biologiska mångfald (in swedish)