Building Corpora for Single-Channel Speech Separation Across Multiple Domains
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2018
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866929564549120000 |
|---|---|
| author | Maciejewski, Matthew Sell, Gregory Garcia-Perera, Leibny Paola Watanabe, Shinji Khudanpur, Sanjeev |
| author_facet | Maciejewski, Matthew Sell, Gregory Garcia-Perera, Leibny Paola Watanabe, Shinji Khudanpur, Sanjeev |
| contents | To date, the bulk of research on single-channel speech separation has been conducted using clean, near-field, read speech, which is not representative of many modern applications. In this work, we develop a procedure for constructing high-quality synthetic overlap datasets, necessary for most deep learning-based separation frameworks. We produced datasets that are more representative of realistic applications using the CHiME-5 and Mixer 6 corpora and evaluate standard methods on this data to demonstrate the shortcomings of current source-separation performance. We also demonstrate the value of a wide variety of data in training robust models that generalize well to multiple conditions. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_1811_02641 |
| institution | arXiv |
| publishDate | 2018 |
| record_format | arxiv |
| spellingShingle | Building Corpora for Single-Channel Speech Separation Across Multiple Domains Maciejewski, Matthew Sell, Gregory Garcia-Perera, Leibny Paola Watanabe, Shinji Khudanpur, Sanjeev Computation and Language To date, the bulk of research on single-channel speech separation has been conducted using clean, near-field, read speech, which is not representative of many modern applications. In this work, we develop a procedure for constructing high-quality synthetic overlap datasets, necessary for most deep learning-based separation frameworks. We produced datasets that are more representative of realistic applications using the CHiME-5 and Mixer 6 corpora and evaluate standard methods on this data to demonstrate the shortcomings of current source-separation performance. We also demonstrate the value of a wide variety of data in training robust models that generalize well to multiple conditions. |
| title | Building Corpora for Single-Channel Speech Separation Across Multiple Domains |
| topic | Computation and Language |
| url | https://arxiv.org/abs/1811.02641 |