Researchers from the United States and Mexico have developed an AI model that analyzes and classifies dog barks, shedding light on their meanings and emotions. This model, named Wav2Vec2, was adapted by Artem Abzaliev, a PhD student at the University of Michigan, in collaboration with scientists from the National Institute of Astrophysics, Optics and Electronics (INAOE) in Puebla.
The Wav2Vec2 model, originally trained to recognize patterns in human speech, was fine-tuned using a dataset of canine vocalizations collected by Humberto Pérez Espinosa, a researcher at INAOE. The recordings, made between 2015 and 2017, include barks from over 100 dogs in various emotional contexts. Stimuli used to provoke canine responses included unexpected noises and simulations of frustration, as well as positive interactions like petting or new toys.
Researchers achieved a 70% accuracy in classifying characteristics such as breed, sex, age, and the context of the bark, surpassing previous systems. Pérez Espinosa noted that the algorithm also allowed for precise differentiation of individual vocalizations. The study highlights that barks vary not only with the dog's emotion but also in intensity, posing challenges for more detailed classifications.
In addition to vocalization analysis, the team is developing multimodal methods that integrate audio, video, movement, and biometric parameters like body temperature and heart rate. These techniques, funded by the National Council of Humanities, Sciences, and Technologies (CONAHCYT), aim to enhance the understanding of canine behavior and have practical applications in training assistance and search dogs, especially in natural disaster scenarios.
One major limitation identified by researchers is the need for sufficient data collection, which requires time, effort, and ethical protocols to ensure animal welfare. Pérez Espinosa emphasized that the project adheres to strict guidelines to guarantee the safety and comfort of the participating dogs.
Moreover, the study aims not only to improve the quality of life for pets through a better understanding of their emotional needs but also to benefit trainers, veterinarians, and animal caregivers. The goal is to generate knowledge with a real social impact, concluded Pérez Espinosa. With these advancements, AI promises to open new frontiers in the study of animal behavior and strengthen the bond between humans and their canine companions.