ChildAugment: Data augmentation methods for zero-resource children's speaker verification
The Journal of the Acoustical Society of America · 01.03.2024 · 155 (3) · 2221-2232
DOI: 10.1121/10.0025178 ↗Kurz gefasst
Fachartikel in der Zeitschrift The Journal of the Acoustical Society of America, online erschienen am 26.03.2024. Verfasst von Vishwanath Pratap Singh und 2 weiteren Autor:innen. Beteiligte Einrichtung: School of Computing, University of Eastern Finland, Joensuu 80130, Finland. Hauptthemen laut MEDLINE: Speech, Acoustics. Die Arbeit wurde bisher 7-mal zitiert (Crossref) und nennt 52 Literaturangaben. Gefördert von: Academy of Finland.
Themen: Kinderkrankenpflege
Abstract (Originalsprache)
The accuracy of modern automatic speaker verification (ASV) systems, when trained exclusively on adult data, drops substantially when applied to children's speech. The scarcity of children's speech corpora hinders fine-tuning ASV systems for children's speech. Hence, there is a timely need to explore more effective ways of reusing adults' speech data. One promising approach is to align vocal-tract parameters between adults and children through children-specific data augmentation, referred here to as ChildAugment. Specifically, we modify the formant frequencies and formant bandwidths of adult speech to emulate children's speech. The modified spectra are used to train emphasized channel attention, propagation, and aggregation in time-delay neural network recognizer for children. We compare ChildAugment against various state-of-the-art data augmentation techniques for children's ASV. We also extensively compare different scoring methods, including cosine scoring, probabilistic linear discriminant analysis (PLDA), and neural PLDA. We also propose a low-complexity weighted cosine score for extremely low-resource children ASV. Our findings on the CSLU kids corpus indicate that ChildAugment holds promise as a simple, acoustics-motivated approach, for improving state-of-the-art deep learning based ASV for children. We achieve up to 12.45% (boys) and 11.96% (girls) relative improvement over the baseline. For reproducibility, we provide the evaluation protocols and codes here.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Autor:innen und Institutionen
- School of Computing, University of Eastern Finland, Joensuu 80130, Finland.
- Institute for Advancing Intelligence, TCG CREST, Kolkata, West Bengal 700091, India.
- Academy of Scientific and Innovative Research (AcSIR), Ghaziabad- 201002, India.
ORCID-Profile: Md Sahidullah, Tomi Kinnunen
Förderung
- Academy of Finland
Zitieren
APA
Singh, V. P., Sahidullah, M., & Kinnunen, T. (2024). ChildAugment: Data augmentation methods for zero-resource children's speaker verification. The Journal of the Acoustical Society of America, 155(3), 2221-2232. https://doi.org/10.1121/10.0025178
Vancouver
Singh VP, Sahidullah M, Kinnunen T. ChildAugment: Data augmentation methods for zero-resource children's speaker verification. The Journal of the Acoustical Society of America. 2024;155(3):2221-2232. doi:10.1121/10.0025178
BibTeX
@article{Singh2024,
author = {Singh, Vishwanath Pratap and Sahidullah, Md and Kinnunen, Tomi},
title = {ChildAugment: Data augmentation methods for zero-resource children's speaker verification},
journal = {The Journal of the Acoustical Society of America},
year = {2024},
volume = {155},
number = {3},
pages = {2221--2232},
doi = {10.1121/10.0025178}
}
EUVIMED