VAST : The Virtual Acoustic Space Traveler Dataset

TL;DR

This paper presents VAST, a large-scale virtual acoustic dataset for training sound source localization models, demonstrating that models trained on this dataset generalize well to real-world data and improve upon traditional methods.

Contribution

Introduction of the VAST dataset for virtual acoustic space traveling, enabling effective learning of sound localization mappings from simulated to real environments.

Findings

Models trained on VAST generalize to real data

VAST dataset improves sound localization accuracy

Overcomes limitations of traditional binaural localization methods

Abstract

This paper introduces a new paradigm for sound source lo-calization referred to as virtual acoustic space traveling (VAST) and presents a first dataset designed for this purpose. Existing sound source localization methods are either based on an approximate physical model (physics-driven) or on a specific-purpose calibration set (data-driven). With VAST, the idea is to learn a mapping from audio features to desired audio properties using a massive dataset of simulated room impulse responses. This virtual dataset is designed to be maximally representative of the potential audio scenes that the considered system may be evolving in, while remaining reasonably compact. We show that virtually-learned mappings on this dataset generalize to real data, overcoming some intrinsic limitations of traditional binaural sound localization methods based on time differences of arrival.