Domestic canid vocalizations: situational context prediction

Abstract

Dogs live all around us and while we have little doubt they communicate with one another, our understanding of this communication is lacking. Our work aims to build a system that will be able to use artificial intelligence techniques to autonomously investigate the meaning behind a dog’s vocalization. It will involve collecting data from a set of animals and recording both elicited and candid vocalizations. The data will then be separated into classes using machine learning as well as information gathered during data collection observation. Finally, the classes will be used to train an audio recognition model for real-time classification. This project’s success could lead to new forms of rescue and service animal training as well as provide a basis on which other mammal vocalizations could be studied.