Artificial intelligence (AI) technology has emerged that may determine a dog’s age, gender, breed, and whether the barking sound is playful or aggressive. The researchers claimed that this might make it possible to develop an AI model that may communicate with dogs.
The BBC reported on the seventh (local time) that researchers on the University of Michigan have developed technology to investigate dog barking sounds using AI and discover nuances.
“Advances in AI might be used to revolutionize our understanding of animal communication,” said Rada Michalcea, director of the University of Michigan’s AI Research Institute. “This research builds on the sphere of speech processing to know the nuances of dog barking.” “It opens a brand new window on how a technology might be used.”
AI has made great strides in learning large numbers of human voices and understanding the subtleties of speech. It runs voice recognition software technology that distinguishes nuances of tone, pitch, and accent.
Nonetheless, the researchers identified that no comparable database of dog voices exists. “The issue with animals is that they can’t request and record specific voices like humans can,” he explains.
Subsequently, the researchers examined whether the model built for humans may very well be applied to dog voices.
First, we collected barking, growling, and whimpering sounds from 74 dogs of varied breeds, ages, and genders. This was applied to ‘Wav2Vec2’, an AI model for human speech evaluation.
In consequence, they found that this method was effective. It was revealed that this model not only succeeded in 4 classification tasks, but additionally outperformed other models specifically trained on dog barking data by as much as 70%.
“That is the primary time that human speech technology has been applied to decipher animal communication,” said Michalcea. “The research results show that human speech can function a basis for analyzing and understanding the acoustic patterns of other sounds, reminiscent of animal vocalizations.” “It shows that it could actually be done,” he said.
Moreover, the researchers emphasized that this discovery could have vital implications for animal welfare.
“Understanding the nuances of the assorted sounds animals make can improve the way in which humans interpret and reply to their emotional and physical needs,” he suggested.
The research was presented on the recent Joint International Conference on Computational Linguistics, Language Resources and Evaluation. Mexico’s National Institute of Astrophysics and the Institute of Optics and Electronics also jointly participated on this project.
Reporter Lim Da-jun ydj@aitimes.com
