Logo des Repositoriums
 

Deep Quality-informed Score Normalization for Privacy-friendly Speaker Recognition in unconstrained Environments

Kurzanzeige
dc.contributor.authorNautsch,Andreas
dc.contributor.authorSteen,Søren Trads
dc.contributor.authorBusch,Christoph
dc.contributor.editorBrömme,Arslan
dc.contributor.editorBusch,Christoph
dc.contributor.editorDantcheva,Antitza
dc.contributor.editorRathgeb,Christian
dc.contributor.editorUhl,Andreas
dc.date.accessioned2017-09-26T09:21:00Z
dc.date.available2017-09-26T09:21:00Z
dc.date.issued2017
dc.description.abstractIn scenarios that are ambitious to protect sensitive data in compliance with privacy regulations, conventional score normalization utilizing large proportions of speaker cohort data is not feasible for existing technology, since the entire cohort data would need to be stored on each mobile device. Hence, in this work we motivate score normalization utilizing deep neural networks. Considering unconstrained environments, a quality-informed scheme is proposed, normalizing scores depending on sample quality estimates in terms of completeness and signal degradation by noise. Utilizing the conventional PLDA score, comparison i-vectors, and corresponding quality vectors, we aim at mimicking cohort based score normalization optimizing the Cmin llr discrimination criterion. Examining the I4U data sets for the 2012 NIST SRE, an 8.7% relative gain is yielded in a pooled 55-condition scenario with a corresponding condition-averaged relative gain of 6.2% in terms of Cmin llr . Robustness analyses towards sensitivity regarding unseen conditions are conducted, i.e. when conditions comprising lower quality samples are not available during training.en
dc.identifier.isbn978-3-88579-664-0
dc.identifier.pissn1617-5468
dc.identifier.urihttps://dl.gi.de/handle/20.500.12116/4655
dc.language.isoen
dc.publisherGesellschaft für Informatik, Bonn
dc.relation.ispartofBIOSIG 2017
dc.relation.ispartofseriesLecture Notes in Informatics (LNI) - Proceedings, Volume P-70
dc.subjectspeaker recognition
dc.subjectscore normalization
dc.subjectunconstrained environments
dc.subjectneural networks
dc.subjectdeep learning
dc.titleDeep Quality-informed Score Normalization for Privacy-friendly Speaker Recognition in unconstrained Environmentsen
gi.citation.endPage250
gi.citation.startPage243
gi.conference.date20.-22. September 2017
gi.conference.locationDarmstadt, Germany
gi.conference.sessiontitleFurther Conference Contributions

Dateien

Originalbündel
1 - 1 von 1
Vorschaubild
Name:
paper24.pdf
Größe:
125.58 KB
Format:
Adobe Portable Document Format
Herunterladen

Sammlungen

P270 - BIOSIG 2017 - Proceedings of the 16th International Conference of the Biometrics Special Interest Group