Deep Normalization for Speaker Vectors

Cai, Yunqi; Li, Lantian; Wang, Dong; Abel, Andrew

doi:10.48550/arxiv.2004.04095

Cited by 7 publications

(15 citation statements)

References 51 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For LDA, the best performance is obtained when the dimension is reduced to 200. For subspace DNF, we can only reduce the dimension to 400, and most performance improvement is not due to dimension reduction, but the nonlinear normalization provided by the full-dimension DNF [13]. Nonetheless, the results show that a nonlinear dimension reduction can be achieved by subspace DNF, and it may lead to additional performance gain over nonlinear normalization.…”

Section: Data and Settingmentioning

confidence: 94%

“…To solve this problem, we extended NF to DNF, a discriminative normalization flow [13]. It is simply a NF but each class owns its individual class mean:…”

Section: Dnf: Real Deep Ldamentioning

confidence: 99%

“…For more details of the MAF architecture, please refer to [17], or the code we published online 1 . More details of the experimental system can be found in our previous work [13]. Table 1 shows the results in terms of Equal Error Rate (EER).…”

Section: Data and Settingmentioning

confidence: 99%

“…Recently, we proposed a discriminative normalization flow (DNF) model and used it to normalize class distributions [13]. In this paper, we reinterpret DNF as a nonlinear extension of LDA.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Deep generative LDA

Cai,

Wang

2020

Preprint

Self Cite

View full text Add to dashboard Cite

Linear discriminant analysis (LDA) is a popular tool for classification and dimension reduction. Limited by its linear form and the underlying Gaussian assumption, however, LDA is not applicable in situations where the data distribution is complex. Recently, we proposed a discriminative normalization flow (DNF) model. In this study, we reinterpret DNF as a deep generative LDA model, and study its properties in representing complex data. We conducted a simulation experiment and a speaker recognition experiment. The results show that DNF and its subspace version are much more powerful than the conventional LDA in modeling complex data and retrieving low-dimensional representations.

show abstract

Section: Data and Settingmentioning

confidence: 94%

“…To solve this problem, we extended NF to DNF, a discriminative normalization flow [13]. It is simply a NF but each class owns its individual class mean:…”

Section: Dnf: Real Deep Ldamentioning

confidence: 99%

Section: Data and Settingmentioning

confidence: 99%

“…Recently, we proposed a discriminative normalization flow (DNF) model and used it to normalize class distributions [13]. In this paper, we reinterpret DNF as a nonlinear extension of LDA.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Deep generative LDA

Cai,

Wang

2020

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…This is not a good property for downstream applications that require discrimination within the latent space, e.g., pronunciation proficiency. Recently, the authors proposed a discriminative NF (DNF) model to deal with this problem [26]. The main advantage of DNF is that it allows each class to have its own Gaussian prior, i.e.…”

Section: Discriminative Nfmentioning

confidence: 99%

ASR-Free Pronunciation Assessment

Cao

Liu²,

et al. 2020

Interspeech 2020

Self Cite

View full text Add to dashboard Cite

Most of the pronunciation assessment methods are based on local features derived from automatic speech recognition (ASR), e.g., the Goodness of Pronunciation (GOP) score. In this paper, we investigate an ASR-free scoring approach that is derived from the marginal distribution of raw speech signals. The hypothesis is that even if we have no knowledge of the language (so cannot recognize the phones/words), we can still tell how good a pronunciation is, by comparatively listening to some speech data from the target language. Our analysis shows that this new scoring approach provides an interesting correction for the phone-competition problem of GOP. Experimental results on the ERJ dataset demonstrated that combining the ASR-free score and GOP can achieve better performance than the GOP baseline.

show abstract

A simulation study on optimal scores for speaker recognition

Wang

2020

J AUDIO SPEECH MUSIC PROC.

Self Cite

View full text Add to dashboard Cite

In this article, we conduct a comprehensive simulation study for the optimal scores of speaker recognition systems that are based on speaker embedding. For that purpose, we first revisit the optimal scores for the speaker identification (SI) task and the speaker verification (SV) task in the sense of minimum Bayes risk (MBR) and show that the optimal scores for the two tasks can be formulated as a single form of normalized likelihood (NL). We show that when the underlying model is linear Gaussian, the NL score is mathematically equivalent to the PLDA likelihood ratio (LR), and the empirical scores based on cosine distance and Euclidean distance can be seen as approximations of this linear Gaussian NL score under some conditions.Based on the unified NL score, we conducted a comprehensive simulation study to investigate the behavior of the scoring component on both the SI task and SV task, in the case where the distribution of the speaker vectors perfectly matches the assumption of the NL model, as well as the case where some mismatch is involved. Importantly, our simulation is based on the statistics of speaker vectors derived from a practical speaker recognition system, hence reflecting the behavior of the NL scoring in real-life scenarios that are full of imperfection, including non-Gaussianality, non-homogeneity, and domain/condition mismatch.

show abstract

Deep Normalization for Speaker Vectors

Cited by 7 publications

References 51 publications

Deep generative LDA

Deep generative LDA

ASR-Free Pronunciation Assessment

A simulation study on optimal scores for speaker recognition

Contact Info

Product

Resources

About