Benchmarking Representations for Speech, Music, and Acoustic Events
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2024
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866914949774704640 |
|---|---|
| author | La Quatra, Moreno Koudounas, Alkis Vaiani, Lorenzo Baralis, Elena Cagliero, Luca Garza, Paolo Siniscalchi, Sabato Marco |
| author_facet | La Quatra, Moreno Koudounas, Alkis Vaiani, Lorenzo Baralis, Elena Cagliero, Luca Garza, Paolo Siniscalchi, Sabato Marco |
| contents | 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 ability to readily incorporate new datasets and models. To address the current lack of open-source, pre-trained models for non-speech audio, we also release new pre-trained models that demonstrate strong performance on non-speech datasets. We argue that the presented wide-ranging evaluation provides valuable insights into state-of-the-art ARL methods, and is useful to pinpoint promising research directions. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2405_00934 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Benchmarking Representations for Speech, Music, and Acoustic Events La Quatra, Moreno Koudounas, Alkis Vaiani, Lorenzo Baralis, Elena Cagliero, Luca Garza, Paolo Siniscalchi, Sabato Marco Audio and Speech Processing Machine Learning Sound 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 ability to readily incorporate new datasets and models. To address the current lack of open-source, pre-trained models for non-speech audio, we also release new pre-trained models that demonstrate strong performance on non-speech datasets. We argue that the presented wide-ranging evaluation provides valuable insights into state-of-the-art ARL methods, and is useful to pinpoint promising research directions. |
| title | Benchmarking Representations for Speech, Music, and Acoustic Events |
| topic | Audio and Speech Processing Machine Learning Sound |
| url | https://arxiv.org/abs/2405.00934 |