aDCF loss function for deep metric learning in end-to-end text-dependent speaker verification systems

Mingote, V. (Universidad de Zaragoza) ; Miguel, A. (Universidad de Zaragoza) ; Ribas, D. ; Ortega, A. (Universidad de Zaragoza) ; Lleida, E. (Universidad de Zaragoza)
aDCF loss function for deep metric learning in end-to-end text-dependent speaker verification systems
Financiación H2020 / H2020 Funds
Resumen: Metric learning approaches have widely expanded to the training of Speaker Verification (SV) systems based on Deep Neural Networks (DNNs), by using a loss function more consistent with the evaluation process than the traditional identification losses. However, these methods do not consider the performance measure and can involve high computational cost, for example, the need for a careful pair or triplet data selection. This paper proposes the approximated Detection Cost Function (aDCF) loss, which is a loss function based on the measure of the decision errors in SV systems, namely the False Rejection Rate (FRR) and the False Acceptance Rate (FAR). With aDCF loss as the training objective function, the end-to-end system learns how to minimize decision errors. Furthermore, we replace the typical linear layer as the last layer of DNN by a cosine distance layer, which reduces the difference between the metric in the training process and the metric during evaluation. aDCF loss function was evaluated in RSR2015-Part I and RSR2015-Part II datasets for text-dependent speaker verification. The system trained with aDCF loss outperforms all the state-of-the-art functions employed in this paper in both parts of the database.
Idioma: Inglés
DOI: 10.1109/TASLP.2022.3145307
Año: 2022
Publicado en: IEEE/ACM Transactions on Audio, Speech, and Language Processing 30 (2022), 772-784
ISSN: 2329-9290

Factor impacto JCR: 5.4 (2022)
Categ. JCR: ENGINEERING, ELECTRICAL & ELECTRONIC rank: 61 / 274 = 0.223 (2022) - Q1 - T1
Categ. JCR: ACOUSTICS rank: 3 / 31 = 0.097 (2022) - Q1 - T1

Factor impacto CITESCORE: 10.1 - Engineering (Q1) - Mathematics (Q1) - Computer Science (Q1) - Physics and Astronomy (Q1)

Factor impacto SCIMAGO: 1.348 - Acoustics and Ultrasonics (Q1) - Computational Mathematics (Q1) - Computer Science (miscellaneous) (Q1) - Speech and Hearing (Q1) - Instrumentation (Q1) - Media Technology (Q1) - Signal Processing (Q1) - Electrical and Electronic Engineering (Q1)

Financiación: info:eu-repo/grantAgreement/ES/AEI/PDC2021-120846-C41
Financiación: info:eu-repo/grantAgreement/ES/DGA/T36-20R
Financiación: info:eu-repo/grantAgreement/EC/H2020/101007666/EU/Exchanges for SPEech ReseArch aNd TechnOlogies/ESPERANTO
Financiación: info:eu-repo/grantAgreement/ES/MCIN/AEI/10.13039/501100011033
Financiación: info:eu-repo/grantAgreement/ES/MINECO/PRE2018-083312
Tipo y forma: Article (PostPrint)
Área (Departamento): Área Teoría Señal y Comunicac. (Dpto. Ingeniería Electrón.Com.)

Rights Reserved All rights reserved by journal editor

Exportado de SIDERAL (2023-09-13-11:09:30)

Este artículo se encuentra en las siguientes colecciones:

 Record created 2022-03-01, last modified 2023-09-14

Rate this document:

Rate this document:
(Not yet reviewed)