Lion roars decoded: AI spots new cat calls

New research utilizes machine learning to identify lion conversations and distinguish each big cat by its unique voice.

Scientists have discovered that African lions produce two different types of roar, which can help identify individual lions and play an important role in conservation efforts.

Using machine learning techniques, a team of researchers from the University of Exeter was able to identify a newly discovered “intermediary roar.”

While it’s not entirely clear what purpose this second roar might serve, the research suggests that it is crucial for understanding lion populations and monitoring individual behavior.

The study, led by third-year Exeter PhD student and conservation technologist Jonathan Growcott, relied on ten continuous days of recordings taken from lions in the Nyerere National Park in southern Tanzania.

Machine learning used to decode roars

In the study, scientists opted for classic, data-efficient machine learning tools rather than deep-learning models, which would have required huge datasets to decode lion roars.

The research paper documents how they measured the length and pitch of each call, then used a clustering model to automatically group roars into categories.

Next, they applied a speech recognition tool called Hidden Markov Models to analyze how the pitch changed over time and confirm the call type.

According to Growcott, spotting such nuances typically relies on expert judgement, but this can be susceptible to human bias. He claims that the AI systems used not only noticed two markedly different types of roar but also identified each individual lion with “a 95% accuracy rate.”

Growcott added: “Lion roars are not just iconic – they are unique signatures that can be used to estimate population sizes and monitor individual animals.”

“Until now, identifying these roars relied heavily on expert judgment, introducing potential human bias. Our new approach using AI promises more accurate and less subjective monitoring, which is crucial for conservationists working to protect dwindling lion populations.”

Why bioacoustics is key in conservation

According to the International Union for Conservation of Nature red list, lions are listed as vulnerable to extinction. The total population of wild lions in Africa is estimated to be between 20,000 and 25,000, but this number has decreased by half in the last 25 years.

Scientists have long attempted to decode animal communication, and Growcott claims that the emerging field of bioacoustics (the study of animal sounds) can be used for counting, monitoring, and protecting this endangered species.

He told BBC’s Today programme: “Traditionally, if you want to do an estimate of lion populations, you would use a camera trap, positioned at trees as lions go past. Bioacoustics means you can detect species like lions much further than a camera. And if you can identify a lion by its roar, you can use that to count the number of individual lions in a landscape.”

“This means you can identify population trends and conservation hotspots which you need to target for protection services.”

The research was a collaborative effort between the University of Exeter, the Wildlife Conservation Unit at the University of Oxford, Lion Landscapes, the Frankfurt Zoological Society, TAWIRI (Tanzania Wildlife Institute for Research), TANAPA (Tanzania National Parks Authority), and computer scientists from Exeter and Oxford.