I want to talk to my Cat (subject 0: Bubbles) Feline Kin and Nurturing Understanding within the Inner Prompting

This paper presents a foray into bioacoustic research at home, focusing on our domestic companion, the cat (Felis catus). The research is conducted through fieldwork, involving audio data collection and processing from a recording device attached to one cat subject, named Bubbles. The aim is to collect cat vocalizations by marking their occurrence with a hotword-marker. Subsequently, hotword-detection speech-to-text models (e.g. Whisper) allow us to extend dataset of animal vocalizations. With such an extended dataset at hand, meow-classifier is trained by means of classical ensemble learning (Random Forest) technique. The paper offers a comparison of these methods and establishes a minimally invasive protocol for future data collection of in vivo and in situ vocalizations of domestic animals.

Read on ResearchGate →