87597 20200729203743.0 doi 10.1186/s13636-019-0167-7 sideral 115809 ART-2019-115809 eng (orcid)0000-0003-1772-0605 Viñals, Ignacio Universidad de Zaragoza Unsupervised adaptation of PLDA models for broadcast diarization 2019 Access copy available to the general public Unrestricted We present a novel model adaptation approach to deal with data variability for speaker diarization in a broadcast environment. Expensive human annotated data can be used to mitigate the domain mismatch by means of supervised model adaptation approaches. By contrast, we propose an unsupervised adaptation method which does not need for in-domain labeled data but only the recording that we are diarizing. We rely on an inner adaptation block which combines Agglomerative Hierarchical Clustering (AHC) and Mean-Shift (MS) clustering techniques with a Fully Bayesian Probabilistic Linear Discriminant Analysis (PLDA) to produce pseudo-speaker labels suitable for model adaptation. We propose multiple adaptation approaches based on this basic block, including unsupervised and semi-supervised. Our proposed solutions, analyzed with the Multi-Genre Broadcast 2015 (MGB) dataset, reported significant improvements (16% relative improvement) with respect to the baseline, also outperforming a supervised adaptation proposal with low resources (9% relative improvement). Furthermore, our proposed unsupervised adaptation is totally compatible with a supervised one. The joint use of both adaptation techniques (supervised and unsupervised) shows a 13% relative improvement with respect to only considering the supervised adaptation. info:eu-repo/grantAgreement/ES/DGA-FEDER/T36-17R info:eu-repo/grantAgreement/ES/MINECO/TIN2017-85854-C4-1-R info:eu-repo/semantics/openAccess by http://creativecommons.org/licenses/by/3.0/es/ 1.289 2019 ENGINEERING, ELECTRICAL & ELECTRONIC 201 / 266 = 0.756 2019 Q4 T3 ACOUSTICS 21 / 32 = 0.656 2019 Q3 T2 0.289 2019 Electrical and Electronic Engineering 2019 Q3 Acoustics and Ultrasonics 2019 Q3 info:eu-repo/semantics/article info:eu-repo/semantics/publishedVersion (orcid)0000-0002-3886-7748 Ortega, Alfonso Universidad de Zaragoza Villalba, Jesús (orcid)0000-0001-5803-4316 Miguel, Antonio Universidad de Zaragoza (orcid)0000-0001-9137-4013 Lleida, Eduardo Universidad de Zaragoza 5008 800 Universidad de Zaragoza Dpto. Ingeniería Electrón.Com. Área Teoría Señal y Comunicac. 2019, 24 (2019), [13 pp.] EURASIP j. audio, speech music. process. EURASIP JOURNAL ON AUDIO SPEECH AND MUSIC PROCESSING 1687-4714 1579181 http://zaguan.unizar.es/record/87597/files/texto_completo.pdf Versión publicada 12113 http://zaguan.unizar.es/record/87597/files/texto_completo.jpg?subformat=icon icon Versión publicada oai:zaguan.unizar.es:87597 articulos driver 2020-07-29-20:20:55 ARTICLE