Can AI interpret pet behaviour based on sound or video ?

How can we decode what animals are 'saying' through their sounds or movements? While technology can now label animal calls or track their body language with reasonable accuracy, turning those observations into clear interpretations of emotion or intent remains a challenge. Current tools exist, but their practical reliability is still in question.

Background

Current systems classify animal vocalizations (e.g., dog barks, cat meows) into broad categories with accuracies ranging from 70% to 90%, varying by species and dataset; however, translating these labels into meaningful emotional or intentional states remains unreliable (Tufts University, 2026). Video-based pose estimation enables real-time tracking of animal movement across multiple joints, yet linking body posture or facial expressions to specific feelings or actions remains a research problem rather than a production capability. Consumer-grade 'bark translators' are offered by start-ups and academic labs, but results are largely anecdotal and lack clinical validation. In welfare science, machine learning is used to detect distress calls in livestock barns, though adoption outside niche applications remains limited.