top of page


There's a tool out there developed at the University of Michigan that can tell a playful bark from an aggressive one . The AI tool can also get information such as the dogs age, breed and sex . A collaboration with Mexico's National Institute of Astrophysics, Optics and Electronic (INAOE) Institute in Puebla. Found AI models trained on human speech can be used as a starting point to train systems that target animal communication.

With voice-enabled technologies they use today such as voice to text and language translation. the models distinguish in human speech tone, pitch and accent and convert this information into a format on a computer, which can identify what words are said and reconise the person speaking.

The incredibly complex patterns of human lanuage and speech are encoded and learnt by these models.

Using a data base of 74 dog vocalizations recorded from different breeds, ages and sex. They wanted to interpret the dogs bark. With a speech representation model calle dWav2Vec2 which was originally trained on human speech data. The model was able to generate representations of the acoustic data collected from dogs. It succeeded at four classifictation tasks and outprefo

rmed other models trained specifically on dog bark data with 70 % accuracy.

The results show sounds and patterns derived from human speech can be a foundation for analyzing and understanding patterns of other sounds , like animal vocalizations.

We are still learning and dont know a lot about the animals we share this planet with. This could help us understand some of animal communication, which could be helpful in animal welfare, how to respond to emotional and physical needs and preventing dangerous situations.

Data and information from Michigan News University of Michigan read written by Emily Fance

3 views0 comments


bottom of page