Researchers Use Artificial Intelligence Models to Decipher Dog Barks
Researchers from the University of Michigan, the National Institute of Astrophysics Mexico, and the Optics and Electronics Institute are utilizing artificial intelligence models to decode the barks of dogs. A new study presented at an international conference showcases how modern AI models can aid in understanding animal communication.
Rada Mikhalch, the director of the Laboratory of AI at the University of Michigan, highlighted the importance of this study: “We do not know much about the animals with which we share this world. Advancements in AI can transform our comprehension of animal communication.”
The study employed the advanced speech recognition model WAV2VEC2, which can determine emotions, gender, and breed of dogs based on their barking. Two different data sets were used for training and comparison: one trained solely on dog barks and the other on human speech modified to sound like barking.
The model, initially trained on 1000 hours of human speech recordings, yielded the best results. After fine-tuning on a data set containing vocalizations of 74 dogs (42 Chihuahuas, 21 French poodles, and 11 schnauzers), it was able to identify emotions with 62% accuracy, breed with 62% accuracy, gender with 69% accuracy, and a specific dog within the group with 50% accuracy.
These outcomes surpass the performance of a model trained exclusively on dog barks, indicating that cues and patterns found in human speech can serve as the foundation for understanding animal communication.
Prior studies demonstrate that the vocalizations of monkeys and meadow dogs can be predicted based on context. Researchers propose that dog barking is similarly linked to context.
The study categorized aggressive barking, regular barking, negative screeching, and grunts as expressions of dog emotions. While dogs experience a wide range of emotions, these sounds were selected for the dataset.
Mikhalch remarked, “By utilizing speech processing models originally trained on human speech, our study unveils a new avenue for leveraging existing technologies to decipher the intricacies of dog barking.”