Compensating noise and reverberation in far-field Multichannel Speaker Verification - Inria - Institut national de recherche en sciences et technologies du numérique Accéder directement au contenu
Pré-Publication, Document De Travail Année : 2022

Compensating noise and reverberation in far-field Multichannel Speaker Verification

Résumé

Speaker verification (SV) suffers from unsatisfactory performance in far-field scenarios due to environmental noise and the adverse impact of room reverberation. This paper investigates utilizing a multichannel pre-processing pipeline including a time-domain neural beamformer (FaSNet), multichannel Wiener filter (MWF), and weighted prediction error (WPE). This approach is compared to the existing state-of-theart approaches. We examine the importance of enrollment in pre-processing which has been largely overlooked in previous studies. Experimental evaluation shows that pre-processing can improve the SV performance as long as the enrollment files are processed similarly to the test data and that test and enrollment occur within similar SNR ranges. The integration of FaSNet, MWF, and WPE achieved improved performance compared to the existing state-of-the-art pre-processing approaches. We also show that our approach generalizes to unseen real recorded data while being trained on simulated data.
Fichier principal
Vignette du fichier
SUBMITTED_VERSION.pdf (222.38 Ko) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-03619903 , version 1 (25-03-2022)
hal-03619903 , version 2 (14-10-2022)

Identifiants

  • HAL Id : hal-03619903 , version 1

Citer

Sandipana Dowerah, Romain Serizel, Denis Jouvet, Mohammad Mohammadamini, Driss Matrouf. Compensating noise and reverberation in far-field Multichannel Speaker Verification. 2022. ⟨hal-03619903v1⟩
206 Consultations
227 Téléchargements

Partager

Gmail Facebook X LinkedIn More