Music source separation via hybrid waveform and spectrogram based generative adversarial network.
Saved in:
| Title: | Music source separation via hybrid waveform and spectrogram based generative adversarial network. |
|---|---|
| Authors: | Wu, Qiuxia1 (AUTHOR) qxwu@scut.edu.cn, Deng, Haipeng1 (AUTHOR) mehpdeng@mail.scut.edu.cn, Hu, Kun2 (AUTHOR) kuhu6123@uni.sydney.edu.au, Wang, Zhiyong2 (AUTHOR) zhiyong.wang@sydney.edu.au |
| Source: | Multimedia Tools & Applications. Sep2025, Vol. 84 Issue 31, p37655-37669. 15p. |
| Subjects: | Generative adversarial networks, Spectrograms, Sound waves, Signal separation, Wave analysis, Loss functions (Statistics), Auditory scene analysis |
| Abstract: | Music source separation aims to disentangle individual sources from the mixture of musical signals. Existing generative adversarial network (GAN) based methods generally work on the spectrogram domain only. However, this practice ignores the patterns from the waveform domain, which are more informative for modelling some categories of sources. In this paper, we propose a fully hybrid GAN framework to integrate knowledge from both domains. In particular, the generator formulates acoustical patterns from waveform and spectrogram domains, while the discriminator provides discriminative information based on the local patch-level spectrograms such that the generator can produce more plausible separation results. Furthermore, to enhance the quality of estimated sources, we devise a perceptual spectrogram loss term, which is a complement of the waveform-level loss. The proposed method is evaluated on two widely used music source separation datasets, producing music sources of high signal-to-distortion ratio (12.03 in MIR-1K dataset and 8.08 in MUSDB18 dataset). These results demonstrate the superiority of the proposed method compared with the state-of-the-art methods. [ABSTRACT FROM AUTHOR] |
| Copyright of Multimedia Tools & Applications is the property of Springer Nature and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.) | |
| Database: | Engineering Source |
|
Full text is not displayed to guests.
Login for full access.
|
|
| FullText | Links: – Type: pdflink Text: Availability: 1 |
|---|---|
| Header | DbId: egs DbLabel: Engineering Source An: 188021219 AccessLevel: 6 PubType: Academic Journal PubTypeId: academicJournal PreciseRelevancyScore: 0 |
| IllustrationInfo | |
| Items | – Name: Title Label: Title Group: Ti Data: Music source separation via hybrid waveform and spectrogram based generative adversarial network. – Name: Author Label: Authors Group: Au Data: <searchLink fieldCode="AR" term="%22Wu%2C+Qiuxia%22">Wu, Qiuxia</searchLink><relatesTo>1</relatesTo> (AUTHOR)<i> qxwu@scut.edu.cn</i><br /><searchLink fieldCode="AR" term="%22Deng%2C+Haipeng%22">Deng, Haipeng</searchLink><relatesTo>1</relatesTo> (AUTHOR)<i> mehpdeng@mail.scut.edu.cn</i><br /><searchLink fieldCode="AR" term="%22Hu%2C+Kun%22">Hu, Kun</searchLink><relatesTo>2</relatesTo> (AUTHOR)<i> kuhu6123@uni.sydney.edu.au</i><br /><searchLink fieldCode="AR" term="%22Wang%2C+Zhiyong%22">Wang, Zhiyong</searchLink><relatesTo>2</relatesTo> (AUTHOR)<i> zhiyong.wang@sydney.edu.au</i> – Name: TitleSource Label: Source Group: Src Data: <searchLink fieldCode="JN" term="%22Multimedia+Tools+%26+Applications%22">Multimedia Tools & Applications</searchLink>. Sep2025, Vol. 84 Issue 31, p37655-37669. 15p. – Name: Subject Label: Subjects Group: Su Data: <searchLink fieldCode="DE" term="%22Generative+adversarial+networks%22">Generative adversarial networks</searchLink><br /><searchLink fieldCode="DE" term="%22Spectrograms%22">Spectrograms</searchLink><br /><searchLink fieldCode="DE" term="%22Sound+waves%22">Sound waves</searchLink><br /><searchLink fieldCode="DE" term="%22Signal+separation%22">Signal separation</searchLink><br /><searchLink fieldCode="DE" term="%22Wave+analysis%22">Wave analysis</searchLink><br /><searchLink fieldCode="DE" term="%22Loss+functions+%28Statistics%29%22">Loss functions (Statistics)</searchLink><br /><searchLink fieldCode="DE" term="%22Auditory+scene+analysis%22">Auditory scene analysis</searchLink> – Name: Abstract Label: Abstract Group: Ab Data: Music source separation aims to disentangle individual sources from the mixture of musical signals. Existing generative adversarial network (GAN) based methods generally work on the spectrogram domain only. However, this practice ignores the patterns from the waveform domain, which are more informative for modelling some categories of sources. In this paper, we propose a fully hybrid GAN framework to integrate knowledge from both domains. In particular, the generator formulates acoustical patterns from waveform and spectrogram domains, while the discriminator provides discriminative information based on the local patch-level spectrograms such that the generator can produce more plausible separation results. Furthermore, to enhance the quality of estimated sources, we devise a perceptual spectrogram loss term, which is a complement of the waveform-level loss. The proposed method is evaluated on two widely used music source separation datasets, producing music sources of high signal-to-distortion ratio (12.03 in MIR-1K dataset and 8.08 in MUSDB18 dataset). These results demonstrate the superiority of the proposed method compared with the state-of-the-art methods. [ABSTRACT FROM AUTHOR] – Name: AbstractSuppliedCopyright Label: Group: Ab Data: <i>Copyright of Multimedia Tools & Applications is the property of Springer Nature and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract.</i> (Copyright applies to all Abstracts.) |
| PLink | https://search.ebscohost.com/login.aspx?direct=true&site=eds-live&db=egs&AN=188021219 |
| RecordInfo | BibRecord: BibEntity: Identifiers: – Type: doi Value: 10.1007/s11042-024-20038-9 Languages: – Code: eng Text: English PhysicalDescription: Pagination: PageCount: 15 StartPage: 37655 Subjects: – SubjectFull: Generative adversarial networks Type: general – SubjectFull: Spectrograms Type: general – SubjectFull: Sound waves Type: general – SubjectFull: Signal separation Type: general – SubjectFull: Wave analysis Type: general – SubjectFull: Loss functions (Statistics) Type: general – SubjectFull: Auditory scene analysis Type: general Titles: – TitleFull: Music source separation via hybrid waveform and spectrogram based generative adversarial network. Type: main BibRelationships: HasContributorRelationships: – PersonEntity: Name: NameFull: Wu, Qiuxia – PersonEntity: Name: NameFull: Deng, Haipeng – PersonEntity: Name: NameFull: Hu, Kun – PersonEntity: Name: NameFull: Wang, Zhiyong IsPartOfRelationships: – BibEntity: Dates: – D: 25 M: 09 Text: Sep2025 Type: published Y: 2025 Identifiers: – Type: issn-print Value: 13807501 Numbering: – Type: volume Value: 84 – Type: issue Value: 31 Titles: – TitleFull: Multimedia Tools & Applications Type: main |
| ResultId | 1 |
