Benchmarking Representations for Speech, Music, and Acoustic Events

arXiv:2405.00934v1 Announce Type: cross
Abstract: Limited diversity in standardized benchmarks for evaluating audio representation learning (ARL) methods may hinder systematic comparison of current methods' capabilities. We present ARCH, a comprehensive benchmark for evaluating ARL methods on diverse audio classification domains, covering acoustic events, music, and speech. ARCH comprises 12 datasets, that allow us to thoroughly assess pre-trained SSL models of different sizes. ARCH streamlines benchmarking of ARL techniques through its unified access to a wide range of domains and its …

abstract arch arxiv audio benchmark benchmarking benchmarks capabilities classification comparison cs.lg cs.sd current datasets diverse diversity domains eess.as events hinder music representation representation learning speech type