AI models decode dog mood and breed by barking

Researchers have found that AI speech recognition models can decipher a dog's mood, breed, and gender by analyzing barking.

AI and Dogs: A New Approach

Researchers at the University of Michigan and the National Institute of Astrophysics, Optics and Electronics in Mexico have published a study showing that AI models, traditionally trained on human speech, can also be applied to dog barking. They used the Wav2Vec2 speech recognition model and fed it with a dataset of barking and noises from 74 dogs of different breeds. The model was able to generate and interpret acoustic data from the dog sounds to determine how the animal was feeling.

Results and challenges

The bark-enriched WAV2VEC2 model was found to be 70% accurate in determining the dog's mood, as well as breed, age, and gender, and outperformed other models trained on similar data. “Animal vocalizations are logistically much harder to obtain and record,” said Artem Abzaliev, lead author of the study. “They must be passively incorporated into the wild or, in the case of pets, with the permission of the owners.”

Innovative data collection

To gather the data needed to train the model, the researchers exposed dogs to “various stimuli,” such as repeatedly ringing the bell or talking to them lovingly, and recorded the sounds the dogs made. These sounds were then fed into the model for analysis.

Applications and future prospects

By reusing a model that was originally designed to analyze human speech, the researchers were able to overcome the data challenge. “These models can learn and code the incredibly complex patterns of human language and speech,” Abzaliev said. “We wanted to see if we could use this ability to distinguish and interpret dog barks.”

Rada Mihalcea, director of the University of Michigan AI Laboratory, added, “By using speech processing models that were initially trained on human speech, our research opens a new window into how we can harness what we've built so far in speech processing to understand the nuances of dog barking.”

A better relationship with our four-legged friends

The research could benefit biologists and animal behaviourists, and help people better respond to dogs' emotional and physical needs. “Our results show that the sounds and patterns derived from human speech can serve as a basis for analyzing and understanding the acoustic patterns of other sounds, such as animal vocalizations,” says Mihalcea.

With this breakthrough in AI research, we may be able to build a better relationship with our four-legged friends and better understand their needs.

‍