Dogs communicate a lot through sounds, just like humans, but how do you decode their language? Researchers have used advanced voice recognition techniques originally designed for human speech to better understand and classify dog barks. The results show that these techniques achieve much better results than traditional methods.
To the ear, the barks all sound the same. However dogsdogs emits vocalizations to tell us a whole lot of things. If owners often struggle to understand their little companion, a US-Mexican research team is looking to artificial intelligence (AI) to make their lives easier.
Researchers from the University of Michigan work in collaboration with the National Institute ofastrophysicsastrophysics, Optics and Electronics (INAOE) in Mexico, to develop an artificial intelligence tool that could interpret dog vocalizations. It would be able to distinguish playful barks from aggressive growls and even identify characteristics such as the dog’s age, breed and gender based solely on the sounds the animal makes.
For their work, the academics used already existing AI models for recognition and transcriptiontranscription vote. ” By using speech processing models originally trained on human voices, our research opens new perspectives on how we can use what we have so far built in fabricfabric speech processing to begin to understand the nuances of dog barking “, explains Rada Mihalcea, professor of computer science and engineering and director of the AI laboratory at the University of Michigan, in a press release.
Four categories of barking
But when Rada Mihalcea and his colleagues were developing their artificial intelligence software, they faced a dearth of acoustic data on dog vocalizations. They therefore decided to record the barks of 74 dogs of different breeds, ages and sexes and have them analyzed by Wav2Vec2, an automatic speech understanding model developed by Meta.
The results were very encouraging: Wav2Vec2 not only managed to classify these barks into four different categories, but it also outperformed other models specifically trained on dog acoustic data, with an accuracy of up to 70%.
” This is the first time that techniques optimized for human speech have been used to decode animal language. We have discovered that sounds and patterns derived from human speech can serve as a basis for analyzing and understanding the acoustic patterns of other sound sources, such as animal vocalizations “, says Rada Mihalcea.
Although this research work is still in the early stages, it offers a glimpse of a future where we will eventually be able to decode animal language. Good news for biologists, but also for dog parents. Indeed, understanding the nuances of canine vocalizations could greatly improve the way humans interpret and respond to emotional and emotional needs. physicalphysical dogs. And therefore to increase the well-being of man’s best friend.
