Deep Normalization for Speaker Vectors

Deep speaker embedding has demonstrated state-of-the-art performance in speaker recognition tasks. However, one potential issue with this approach is that the speaker vectors derived from deep embedding models tend to be non-Gaussian for each individual speaker, and non-homogeneous for distributions...

Full description

Saved in:
Bibliographic Details
Published inIEEE/ACM transactions on audio, speech, and language processing Vol. 29; pp. 733 - 744
Main Authors Cai, Yunqi, Li, Lantian, Abel, Andrew, Zhu, Xiaoyan, Wang, Dong
Format Journal Article
LanguageEnglish
Published Piscataway IEEE 2021
The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
Subjects
Covariance matrices
Dimensionality reduction
Embedding
Gaussian distribution
Normal distribution
Normalization flow
Probabilistic logic
speaker embedding
Speaker recognition
Speech recognition
Task analysis
Training
Transforms
Online AccessGet full text

Cover

More Information
Summary:Deep speaker embedding has demonstrated state-of-the-art performance in speaker recognition tasks. However, one potential issue with this approach is that the speaker vectors derived from deep embedding models tend to be non-Gaussian for each individual speaker, and non-homogeneous for distributions of different speakers. These irregular distributions can seriously impact speaker recognition performance, especially with the popular PLDA scoring method, which assumes homogeneous Gaussian distribution. In this article, we argue that deep speaker vectors require deep normalization, and propose a deep normalization approach based on a novel discriminative normalization flow (DNF) model. We demonstrate the effectiveness of the proposed approach with experiments using the widely used SITW and CNCeleb corpora. In these experiments, the DNF-based normalization delivered substantial performance gains and also showed strong generalization capability in out-of-domain tests.
ISSN:2329-9290
2329-9304
DOI:10.1109/TASLP.2020.3039573