Thesis Open Access

SATB Voice Segregation For Monoaural Recordings

Pétermann, Darius A,

DataCite XML Export

<?xml version='1.0' encoding='utf-8'?>
<identifier identifierType="DOI">10.5281/zenodo.4091247</identifier>
<creators>
<creator>
<creatorName>Pétermann, Darius A,</creatorName>
<affiliation>Universitat Pompeu Fabra</affiliation>
</creator>
</creators>
<titles>
<title>SATB Voice Segregation For Monoaural Recordings</title>
</titles>
<publisher>Zenodo</publisher>
<publicationYear>2020</publicationYear>
<subjects>
<subject>source separation, singing voice, SATB recording, convolutional neural networks, choir music.</subject>
</subjects>
<contributors>
<contributor contributorType="Supervisor">
<contributorName>Chandna, Pritish</contributorName>
<givenName>Pritish</givenName>
<familyName>Chandna</familyName>
<affiliation>Universitat Pompeu Fabra</affiliation>
</contributor>
<contributor contributorType="Supervisor">
<givenName>Jordi</givenName>
<affiliation>Universitat Pompeu Fabra</affiliation>
</contributor>
</contributors>
<dates>
<date dateType="Issued">2020-09-15</date>
</dates>
<resourceType resourceTypeGeneral="Text">Thesis</resourceType>
<alternateIdentifiers>
<alternateIdentifier alternateIdentifierType="url">https://zenodo.org/record/4091247</alternateIdentifier>
</alternateIdentifiers>
<relatedIdentifiers>
<relatedIdentifier relatedIdentifierType="DOI" relationType="IsVersionOf">10.5281/zenodo.4091246</relatedIdentifier>
<relatedIdentifier relatedIdentifierType="URL" relationType="IsPartOf">https://zenodo.org/communities/mtgupf</relatedIdentifier>
<relatedIdentifier relatedIdentifierType="URL" relationType="IsPartOf">https://zenodo.org/communities/smc-master</relatedIdentifier>
</relatedIdentifiers>
<rightsList>
<rights rightsURI="info:eu-repo/semantics/openAccess">Open Access</rights>
</rightsList>
<descriptions>
<description descriptionType="Abstract">&lt;p&gt;Choral singing is a widely practiced form of ensemble singing wherein a group of people sing simultaneously in polyphonic harmony. The most commonly practiced&amp;nbsp;setting for choir ensembles consists of four parts; Soprano, Alto, Tenor and Bass&amp;nbsp;(SATB), each with its own range of fundamental frequencies (F0s). The task of&amp;nbsp;source separation for this choral setting entails separating the SATB mixture into&amp;nbsp;its constituent parts. Source separation for musical mixtures is well studied and&amp;nbsp;many Deep Learning-based methodologies have been proposed for the same. However,&lt;br&gt;
most of the research has been focused on a typical case which consists in&lt;br&gt;
separating vocal, percussion and bass sources from a mixture, each of which has a&amp;nbsp;distinct spectral structure. In contrast, the simultaneous and harmonic nature of&amp;nbsp;ensemble singing leads to high structural similarity and overlap between the spectral&amp;nbsp;components of the sources in a choral mixture, making source separation for&amp;nbsp;choirs a harder task than the typical case. This, along with the lack of an appropriate&amp;nbsp;consolidated dataset has led to a dearth of research in the field so far. In&amp;nbsp;this work we first assess how well some of the recently developed methodologies for&amp;nbsp;musical source separation perform for the case of SATB choirs. We then propose a&amp;nbsp;novel domain-specific adaptation for conditioning the recently proposed U-Net architecture&lt;br&gt;
for musical source separation using the fundamental frequency contour of&lt;br&gt;
each of the singing groups and demonstrate that our proposed approach surpasses&amp;nbsp;results from domain-agnostic architectures. Lastly we assess our approach using&amp;nbsp;different evaluation methodologies, going from objective to subjective-based ones,&amp;nbsp;and provide a comparative analysis of the various results.&lt;/p&gt;</description>
</descriptions>
</resource>

237
169
views