Cross-lingual Text-independent Speaker Verification using Unsupervised Adversarial Discriminative Domain Adaptation

Jing Huang; John H.L. Hansen; Wei Xia

REVIEW 3 major objections 5 minor 32 references

Cross-lingual Text-independent Speaker Verification using Unsupervised Adversarial Discriminative Domain Adaptation

T0 review · 3 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Unlabeled Chinese speech adapts an English-trained speaker verifier, cutting errors by 18 percent.

desk verdict A credible but unpolished application paper: ADDA on x-vectors shows a promising EER gain, but the abstract/body number mismatch and missing error bars mean the headline should be treated as provisional. read the letter →

arxiv 1908.01447 v1 pith:YAS4JTAT submitted 2019-08-05 eess.AS cs.LGcs.SD

Wei Xia , Jing Huang , John H.L. Hansen This is my paper · ORCID

classification eess.AScs.LGcs.SD

keywords speakerverificationcross-lingualdomainadaptationadversarialdiscriminativex-vectorunsupervisedlearningequalerrorrateAISHELL

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper aims to show that a speaker-verification system trained only on English can be made to work on Chinese without any labeled Chinese speech. It does this by applying Adversarial Discriminative Domain Adaptation (ADDA) to x-vector speaker embeddings: a source encoder trained on English data is kept fixed, and a separate target encoder is trained to fool a domain discriminator so that Chinese x-vectors are mapped into the English embedding space. On the AISHELL cross-lingual task, this unsupervised adaptation lowers the Equal Error Rate from 9.331% to 7.645%, a relative 18.07% reduction, and improves over a Domain Adversarial Training (DAT) baseline by 12.54% relative. If the result holds, robust speaker verification for a new language can be obtained from unlabeled audio alone, avoiding the expense of target-language speaker labels.

What carries the argument

The central mechanism is the adversarial discriminative domain adaptation objective with asymmetric encoders. A source DNN encoder $M_s$ and speaker classifier $C$ are trained on labeled English x-vectors, after which $M_s$ is frozen. A domain discriminator $D$ is trained to distinguish $M_s(x_s)$ from the target encoder output $M_t(x_t)$, while $M_t$ is trained to maximize $\mathbb{E}_{x_t\sim X_t}[\log D(M_t(x_t))]$, giving stronger gradients than the minimax loss used in DAT and avoiding the assumption that source and target domains share a similar class distribution. The x-vector model supplies the base speaker embedding, and a PLDA back-end scores the adapted enrollment and test vectors after mean and length normalization.

What would settle it

Use a different Mandarin corpus, or a different recording condition, as the unlabeled adaptation set while keeping the same evaluation speakers; if the Equal Error Rate improvement over the unadapted x-vector baseline drops well below 18.07% relative, the transfer is driven by the adaptation set matching the test domain rather than by the ADDA mechanism alone.

Watch

Extended reading notes

Core claim

The paper's central claim is that asymmetric adversarial adaptation, rather than a shared feature extractor, is the right way to transfer an English-trained speaker encoder to a new language. The target encoder is initialized from the frozen source encoder and then optimized with the loss $-\mathbb{E}_{x_t\sim X_t}[\log D(M_t(x_t))]$, where $D$ is a discriminator that tries to tell source and target x-vectors apart. The paper reports that this ADDA embedding reduces EER on the Chinese AISHELL evaluation from 9.331% to 7.645% (18.07% relative), and from 8.741% to 7.645% relative to the x-vector plus DAT concatenation (12.54% relative). Clustering analysis backs the claim: Normalized Mutual Information of K-means on the adapted embeddings rises from 0.787 to 0.802, indicating the adapted representations are more speaker-discriminative and less tied to the source language.

Load-bearing premise

The argument depends on the unlabeled AISHELL training data being representative of the held-out AISHELL evaluation speakers in language, recording channel, and speaker demographics; if those conditions diverge, the reported error reduction could shrink or disappear.

Editorial extensions

If this is right

A speaker-verification system can be extended to a new language using only unlabeled audio, eliminating the need for target-language speaker labels.
The separate-encoder ADDA strategy outperforms the shared-encoder DAT baseline on the English-to-Chinese task, suggesting asymmetric adaptation is better suited to large language shifts.
The gain transfers to held-out evaluation speakers, not just to the unlabeled adaptation speakers, so the method is not simply memorizing the adaptation set.
ADDA and unsupervised PLDA adaptation are complementary: combining them lowers EER from 7.645% to 7.504%.
Adapted embeddings are more clusterable in the target language (NMI 0.802 vs 0.787), so language-invariance and speaker-discriminability improve together.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

A natural extension is to apply the same frozen-source and adversarial-target recipe to other language pairs or to channel mismatch; the paper's result suggests the method does not depend on the particular English-Chinese pair, but that is our extrapolation, not its claim.
A testable check would replace the AISHELL unlabeled set with a Mandarin corpus recorded under different conditions; if the EER gain mostly vanishes, transfer comes from the unlabeled set matching the evaluation domain rather than from the adaptation objective itself.
The same ADDA adaptation could be applied to newer neural speaker encoders, though the relative gain may shrink if the base encoder already produces more language-robust embeddings.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Short version: this is a solid empirical application of ADDA to x-vector speaker verification, with a clean held-out evaluation. The headline EER drop from 9.33% to 7.65% is credible, but the paper has an internal inconsistency in the reported improvement over the DAT baseline, and the lack of any variance estimate leaves the size of the gain uncertain. Worth refereeing; needs a revision.

What's new: ADDA itself is a published CVPR method, and DAT had already been tried for i-vector speaker recognition by Wang et al. 2018. The novelty here is the specific test: using ADDA to adapt x-vector embeddings across languages, with English SRE/SWBD as source and Chinese AISHELL as target. The empirical setup is reasonable: separate source and target encoders, an adversarial discriminator trained on unlabeled target x-vectors, and evaluation on held-out AISHELL test trials with 20 speakers, no overlap with the 340 speakers used for adaptation. The DET curve and MinDCF support the EER numbers. The clustering NMI improvement is small but in the right direction. Those are real positives.

Soft spots: The abstract says 6.32% relative EER reduction from DAT, while Section 4.1 says 12.54%. The table numbers give 12.54% (8.741 vs 7.645). That factor-of-two discrepancy must be fixed. More importantly, there are no error bars or significance tests; with only 20 test speakers, a 1.1-point EER difference could be within run-to-run noise. I'd like to see at least a bootstrap confidence interval or multiple random initializations. The weakest assumption is that the unlabeled adaptation data (AISHELL training set) is representative of the test conditions; it's from the same corpus, so the adaptation is somewhat in-domain. That's a limitation but not fatal, and the paper is transparent about it. Also, the claim that ADDA is 'considerably simpler' is only supported against one baseline method.

Who this is for: speech engineers working on cross-lingual or cross-domain speaker verification, and anyone applying adversarial domain adaptation to embeddings. It deserves a serious referee because the combination is new and the result, if robust, is practically useful. I would not desk-reject it, but the final version should be conditional on resolving the number mismatch and adding a measure of uncertainty.

Referee Report

3 major / 5 minor

Summary. The paper proposes applying Adversarial Discriminative Domain Adaptation (ADDA) to x-vector speaker embeddings for cross-lingual text-independent speaker verification. The source domain is English telephone/overlapping speech (NIST SRE04-08, Mixer 6, Switchboard) and the target domain is Mandarin AISHELL-I. The method trains a source DNN encoder with speaker labels, then adapts a separate target encoder adversarially using only unlabeled target x-vectors. Experiments report EER on the AISHELL test set for x-vector baseline, x-vector+DAT (concatenated) baseline, and ADDA, with and without PLDA adaptation. The headline result is that ADDA reduces EER from 9.331% to 7.645% (18.07% relative) and gives a relative improvement over the DAT concatenation baseline. The paper also includes t-SNE visualization and K-means NMI analysis of the adapted embeddings.

Significance. If the reported gains hold, the paper provides a useful and simple application of a known adversarial domain adaptation method to a practical speaker verification problem: adapting an English-trained x-vector system to a new language with unlabeled target data. The evaluation is externally grounded: the AISHELL test set has no speaker overlap with the unlabeled adaptation data, and the method is a prior published approach with no fitted constant hidden in the headline claim. The main value is as a demonstration that unsupervised target-encoder adaptation on x-vector embeddings can reduce cross-lingual EER. However, the central quantitative claim is currently supported only by point estimates with no variance information, and the abstract and body report inconsistent improvement numbers over the DAT baseline. These issues must be resolved before the performance claim can be considered established.

major comments (3)

[Abstract vs. Section 4.1, Table 1] The abstract states that ADDA gives a 6.32% relative reduction from DAT, while Section 4.1 states a 12.54% relative improvement compared with the concatenated x-vector+DAT embedding. From Table 1, (8.741 - 7.645)/8.741 = 12.54%, so the abstract number is inconsistent with the only tabulated DAT-based baseline. This is a factor-of-two discrepancy in the headline result and must be corrected or explained.
[Section 4.1, Table 1] The EER values are reported as point estimates without error bars, multiple seeds, cross-validation splits, or significance tests. The key comparisons are small in absolute terms: ADDA vs. x-vector+DAT differs by 1.096 EER, and with PLDA adaptation by only 0.295 EER. Given that the test set contains only 20 speakers, one or two speaker-level outliers could materially change these numbers, and run-to-run variation of the adversarial training is known to be non-negligible. The authors should provide variance estimates, e.g., multiple random initializations or bootstrap confidence intervals over speakers, to support the relative improvement claims.
[Section 4.1, Table 1] The claimed improvement 'compared with DAT' is actually measured against the concatenated x-vector+DAT embedding, not against a standalone DAT system. No EER for DAT alone is reported, although the text states that concatenation always performs better than a single DAT embedding. Since the abstract and conclusions frame the contribution as being better than DAT, the absence of a standalone DAT result weakens the direct comparison. Reporting the standalone DAT number, or explicitly rephrasing the claim to describe the concatenated baseline, is needed.

minor comments (5)

[Section 3.3] There is a typographical error: 'AISEHLL unlabeled x-vectors' should read 'AISHELL unlabeled x-vectors.'
[Equation (1)] The indexing in the DAT loss is unclear: the first domain-classifier term is scaled by 1/n but sums over the source samples, while the second sums i = n+1 to N with scale 1/n'. Please clarify the relation between n, n', and N, and make the target-sample count explicit.
[Section 3.3] The description 'energy based V AD' should be 'energy-based VAD' with proper spacing.
[Section 4.1, Table 1] The table caption says 'with a PLDA back-end,' but the table includes rows both with and without PLDA adaptation. Please rephrase to indicate that PLDA adaptation is applied only to the last three rows.
[Section 2.3] The notation defining x and y as 'x-vectors and speaker labels' is potentially confusing because x-vectors are already embeddings; consider using terms such as 'input features' or 'source-domain embeddings' to avoid ambiguity.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the central EER claim is an external held-out benchmark with a prior published adaptation method.

full rationale

The paper's central claim is an empirical EER comparison on the AISHELL-I test trials. The chain is: (i) train an x-vector model on English SRE04-08/Mixer6/Switchboard data; (ii) adapt a target encoder with unlabeled AISHELL training data using ADDA; (iii) evaluate the adapted embeddings on AISHELL test enrollment/test trials (143,520 trials, 20 held-out speakers, with no speaker overlap with the adaptation set). ADDA (Tzeng et al., CVPR 2017) and DAT (Ganin et al., JMLR 2016) are prior external methods; the x-vector baseline, PLDA back-end, and AISHELL corpus are also external to this paper. No parameter is fitted to the test set and then reported as a prediction; the EER numbers are measured on held-out trials. The paper's self-citations ([3], [11], [13], [18], [19]) appear only in related-work descriptions and are not load-bearing for the reported gain. The abstract/body discrepancy in the DAT-relative improvement (6.32% vs 12.54%) and the absence of error bars or significance testing are correctness and statistical-reporting concerns, but they do not make the derivation circular. Therefore, no circular step is present.

Assumptions & free parameters 5 free parameters · 5 assumptions · 0 invented entities

The paper's central claim is an empirical performance comparison, so the ledger captures the modeling assumptions and hand-chosen hyperparameters that the result depends on. No new entities are introduced, and the ADDA method itself is prior work.

free parameters (5)

learning_rate = 1e-4
Adam optimizer learning rate chosen by hand for both source DNN training and ADDA adaptation (Section 3.4).
batch_size = 128
Batch size used for all training phases (Section 3.4).
epochs = 100 (source DNN), 100 (ADDA target encoder)
Number of training epochs chosen without stated early stopping or validation (Section 3.4).
LDA_dimension = 256
Embeddings are LDA-reduced to 256 dimensions before scoring and visualization (Section 4.1).
DAT_loss_weight = equal weights (lambda = 1)
Multi-task DAT loss combines speaker and domain losses with equal weights, a hand choice (Section 3.4).

assumptions (5)

domain assumption Unlabeled target-domain speech is available from the same distribution as the evaluation domain
The method uses AISHELL unlabeled training data to adapt the target encoder, and the test set is from the same corpus, so the assumption is that this unlabeled data is representative of the evaluation conditions (Sections 3.2, 3.3).
domain assumption x-vector embeddings capture the information needed for speaker verification
The paper builds on x-vector as the baseline and applies adaptation to these embeddings rather than to the original acoustic model, assuming that the TDNN embeddings retain the speaker-discriminative information needed for the task (Sections 2.1, 3.3).
domain assumption Adversarial domain adaptation can align feature distributions without destroying speaker-discriminative information
This is the premise of both DAT and ADDA; the paper assumes that training the target encoder to fool the discriminator will produce language-invariant but speaker-discriminative embeddings (Sections 2.2, 2.3, Eqs. 3 and 4).
domain assumption PLDA trained on English data generalizes to adapted embeddings
The scoring back-end is trained on English SRE clean and noise-augmented data, then applied without retraining to the adapted Chinese embeddings (Section 3.3).
standard math The ADDA training objective has the same fixed-point properties as the minimax loss but stronger gradients
Section 2.3, Eq. 4 states this property, taken from Tzeng et al. (2017), and uses it to justify the asymmetric mapping. It is an unproved assertion in this paper, inherited from the cited prior work.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Cross-lingual Text-independent Speaker Verification using Unsupervised Adversarial Discriminative Domain Adaptation." pith.science (2026). https://pith.science/paper/YAS4JTAT

@misc{pith2026190801447,
  author       = {Pith},
  title        = {Pith review of: Cross-lingual Text-independent Speaker Verification using Unsupervised Adversarial Discriminative Domain Adaptation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YAS4JTAT}},
  note         = {Machine review of arXiv:1908.01447}
}

read the original abstract

Speaker verification systems often degrade significantly when there is a language mismatch between training and testing data. Being able to improve cross-lingual speaker verification system using unlabeled data can greatly increase the robustness of the system and reduce human labeling costs. In this study, we introduce an unsupervised Adversarial Discriminative Domain Adaptation (ADDA) method to effectively learn an asymmetric mapping that adapts the target domain encoder to the source domain, where the target domain and source domain are speech data from different languages. ADDA, together with a popular Domain Adversarial Training (DAT) approach, are evaluated on a cross-lingual speaker verification task: the training data is in English from NIST SRE04-08, Mixer 6 and Switchboard, and the test data is in Chinese from AISHELL-I. We show that with the ADDA adaptation, Equal Error Rate (EER) of the x-vector system decreases from 9.331\% to 7.645\%, relatively 18.07\% reduction of EER, and 6.32\% reduction from DAT as well. Further data analysis of ADDA adapted speaker embedding shows that the learned speaker embeddings can perform well on speaker classification for the target domain data, and are less dependent with respect to the shift in language.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

32 extracted references · 31 canonical work pages

[1]

Cross-lingual Text-independent Speaker Verification using Unsupervised Adversarial Discriminative Domain Adaptation

INTRODUCTION Speaker veriﬁcation (SV) offers a natural and ﬂexible option for bio- metric authentication. The text-independent SV system, which does not require the ﬁxed input voice content, is a ﬂexible and challenging task. In real-world scenarios, however, speaker veriﬁcation systems may degrade signiﬁcantly when training on one language and test it on...

work page Pith review arXiv 1908
[2]

SPEAKER VERIFICATION SYSTEMS 2.1. The X-vector system We use a recently proposed successful speaker model called X- vector [6], to extract speaker representations, and a Probabilistic Linear Discriminant Analysis (PLDA) back-end to compare pairs of enrollment and test speaker embeddings. The X-vector model is based on a Time-Delay Deep Neural Network (TDN...

work page
[3]

English Corpora We use Speaker Recognition Evaluation (SRE) 04-08, Mixer 6, and Switchboard (SWBD) to train the x-vector model

EXPERIMENTAL SETUP 3.1. English Corpora We use Speaker Recognition Evaluation (SRE) 04-08, Mixer 6, and Switchboard (SWBD) to train the x-vector model. SRE corpus is part of the Mixer 6 project, which was designed to support the devel- opment of robust speaker recognition technology by providing care- fully collected speech across numerous microphones. Sw...

work page
[4]

0764”, it is difﬁcult to separate it from speaker “0765

RESULTS AND DISCUSSIONS 4.1. Results In this section, we show experimental results using x-vector, x- vector with DAT and x-vector with ADDA training with and without PLDA adaptations in Table 1. We use Linear Discriminant Analysis (LDA) to reduce all three embeddings to 256 dimension for compar- ison. Also, we concatenate the DAT embedding with the x-vec...

work page
[5]

CONCLUSIONS AND FUTURE WORK We presented a discriminative adversarial unsupervised adaptation method in this paper. By exploiting how to alleviate the domain mismatch problem in an English-Chinese cross-lingual speaker ver- iﬁcation task, we showed that our proposed unsupervised ADDA approach can perform well on speaker classiﬁcation for the target domain...

work page
[6]

Joint factor analysis versus eigenchannels in speaker recognition,

Patrick Kenny, Gilles Boulianne, Pierre Ouellet, and Pierre Dumouchel, “Joint factor analysis versus eigenchannels in speaker recognition,” IEEE Transactions on Audio, Speech, and Language Processing, vol. 15, no. 4, pp. 1435–1447, 2007

work page 2007
[7]

Full-covariance ubm and heavy-tailed plda in i- vector speaker veriﬁcation,

Pavel Mat ˇejka, Ondˇrej Glembek, Fabio Castaldo, Md Jahangir Alam, Oldˇrich Plchot, Patrick Kenny, Luk ´aˇs Burget, and Jan ˇCernocky, “Full-covariance ubm and heavy-tailed plda in i- vector speaker veriﬁcation,” in Acoustics, Speech and Sig- nal Processing (ICASSP), IEEE International Conference on , 2011, pp. 4828–4831

work page 2011
[8]

Speaker recognition by machines and humans: A tutorial review,

John HL Hansen and Tauﬁq Hasan, “Speaker recognition by machines and humans: A tutorial review,” IEEE Signal pro- cessing magazine, vol. 32, no. 6, pp. 74–99, 2015

work page 2015

Show all 32 references

[9]

Bayesian speaker veriﬁcation with heavy- tailed priors.,

Patrick Kenny, “Bayesian speaker veriﬁcation with heavy- tailed priors.,” in Odyssey, 2010, p. 14

2010
[10]

Probabilistic linear dis- criminant analysis for inferences about identity,

Simon JD Prince and James H Elder, “Probabilistic linear dis- criminant analysis for inferences about identity,” in Computer Vision (ICCV), IEEE International Conference on , 2007, pp. 1–8

2007
[11]

X-vectors: Robust dnn em- beddings for speaker recognition,

David Snyder, Daniel Garcia-Romero, Gregory Sell, Daniel Povey, and Sanjeev Khudanpur, “X-vectors: Robust dnn em- beddings for speaker recognition,” Acoustics, Speech and Sig- nal Processing (ICASSP), IEEE International Conference on , 2018

2018
[12]

Conditional gener- ative adversarial networks for speech enhancement and noise- robust speaker veriﬁcation,

Daniel Michelsanti and Zheng-Hua Tan, “Conditional gener- ative adversarial networks for speech enhancement and noise- robust speaker veriﬁcation,” in INTERSPEECH, 2017

2017
[13]

Attention-based models for text- dependent speaker veriﬁcation,

FA Rezaur rahman Chowdhury, Quan Wang, Ignacio Lopez Moreno, and Li Wan, “Attention-based models for text- dependent speaker veriﬁcation,” in Acoustics, Speech and Sig- nal Processing (ICASSP), IEEE International Conference on , 2018, pp. 5359–5363

2018
[14]

End-to-end attention based text-dependent speaker ver- iﬁcation,

Shi-Xiong Zhang, Zhuo Chen, Yong Zhao, Jinyu Li, and Yifan Gong, “End-to-end attention based text-dependent speaker ver- iﬁcation,” in Spoken Language Technology Workshop (SLT). IEEE, 2016, pp. 171–178

2016
[15]

Generalized end-to-end loss for speaker veriﬁcation,

Li Wan, Quan Wang, Alan Papir, and Ignacio Lopez Moreno, “Generalized end-to-end loss for speaker veriﬁcation,” in Acoustics, Speech and Signal Processing (ICASSP), IEEE In- ternational Conference on, 2018, pp. 4879–4883

2018
[16]

Text-independent speaker veriﬁcation based on triplet convo- lutional neural network embeddings,

Chunlei Zhang, Kazuhito Koishida, and John HL Hansen, “Text-independent speaker veriﬁcation based on triplet convo- lutional neural network embeddings,”IEEE/ACM Transactions on Audio, Speech and Language Processing (TASLP), vol. 26, no. 9, pp. 1633–1644, 2018

2018
[17]

Ad- versarial network bottleneck features for noise robust speaker veriﬁcation,

Hong Yu, Zheng-Hua Tan, Zhanyu Ma, and Jun Guo, “Ad- versarial network bottleneck features for noise robust speaker veriﬁcation,” in INTERSPEECH, 2017, pp. 1492–1496

2017
[18]

Speaker recognition with non- linear distortion: Clipping analysis and impact,

Wei Xia and John HL Hansen, “Speaker recognition with non- linear distortion: Clipping analysis and impact,” in INTER- SPEECH, 2018, pp. 746–750

2018
[19]

End-to-end text-dependent speaker veriﬁcation,

Georg Heigold, Ignacio Moreno, Samy Bengio, and Noam Shazeer, “End-to-end text-dependent speaker veriﬁcation,” in Acoustics, Speech and Signal Processing (ICASSP), IEEE In- ternational Conference on, 2016, pp. 5115–5119

2016
[20]

Joint Training of Expanded End-to-End DNN for Text-Dependent Speaker Veriﬁcation,

Hee-soo Heo, Jee-weon Jung, IL-ho Yang, Sung-hyun Yoon, and Ha-jin Yu, “Joint Training of Expanded End-to-End DNN for Text-Dependent Speaker Veriﬁcation,” inINTERSPEECH, 2017, pp. 1532–1536

2017
[21]

Inter dataset variability compensation for speaker recognition,

Hagai Aronowitz, “Inter dataset variability compensation for speaker recognition,” in Acoustics, Speech and Signal Pro- cessing (ICASSP), IEEE International Conference on , 2014, pp. 4002–4006

2014
[22]

Improving out-domain plda speaker veriﬁcation using unsu- pervised inter-dataset variability compensation approach,

Ahilan Kanagasundaram, David Dean, and Sridha Sridharan, “Improving out-domain plda speaker veriﬁcation using unsu- pervised inter-dataset variability compensation approach,” in Acoustics, Speech and Signal Processing (ICASSP), IEEE In- ternational Conference on, 2015, pp. 4654–4658

2015
[23]

Maximum-likelihood linear transformation for unsupervised domain adaptation in speaker veriﬁcation,

Abhinav Misra and John HL Hansen, “Maximum-likelihood linear transformation for unsupervised domain adaptation in speaker veriﬁcation,” IEEE/ACM Transactions on Audio, Speech and Language Processing (TASLP), vol. 26, no. 9, pp. 1549–1558, 2018

2018
[24]

Modelling and compen- sation for language mismatch in speaker veriﬁcation,

Abhinav Misra and John HL Hansen, “Modelling and compen- sation for language mismatch in speaker veriﬁcation,” Speech Communication, vol. 96, pp. 58–66, 2018

2018
[25]

Domain-adversarial training of neural networks,

Yaroslav Ganin, Evgeniya Ustinova, Hana Ajakan, Pascal Ger- main, Hugo Larochelle, Franc ¸ois Laviolette, Mario Marchand, and Victor Lempitsky, “Domain-adversarial training of neural networks,” The Journal of Machine Learning Research , vol. 17, no. 1, pp. 2096–2030, 2016

2016
[26]

Multi-adversarial domain adaptation,

Zhongyi Pei, Zhangjie Cao, Mingsheng Long, and Jianmin Wang, “Multi-adversarial domain adaptation,” in AAAI Con- ference on Artiﬁcial Intelligence, 2018

2018
[27]

Adversarial deep averaging networks for cross-lingual sentiment classiﬁcation,

Xilun Chen, Yu Sun, Ben Athiwaratkun, Claire Cardie, and Kilian Weinberger, “Adversarial deep averaging networks for cross-lingual sentiment classiﬁcation,” arXiv preprint arXiv:1606.01614, 2016

2016 arXiv
[28]

Unsupervised domain adaptation via domain adversarial training for speaker recognition,

Qing Wang, Wei Rao, Sining Sun, Leib Xie, Eng Siong Chng, and Haizhou Li, “Unsupervised domain adaptation via domain adversarial training for speaker recognition,” in Acoustics, Speech and Signal Processing (ICASSP), IEEE International Conference on, 2018, pp. 4889–4893

2018
[29]

Adversarial discriminative domain adaptation,

Eric Tzeng, Judy Hoffman, Kate Saenko, and Trevor Darrell, “Adversarial discriminative domain adaptation,” in Computer Vision and Pattern Recognition (CVPR), 2017, vol. 1, p. 4

2017
[30]

Aishell-1: An open-source mandarin speech corpus and a speech recognition baseline,

Hui Bu, Jiayu Du, Xingyu Na, Bengu Wu, and Hao Zheng, “Aishell-1: An open-source mandarin speech corpus and a speech recognition baseline,” in Conference of the Orien- tal Chapter of the International Coordinating Committee on Speech Databases and Speech I/O Systems and Assess...

2017
[31]

Visualizing data using t-sne,

Laurens van der Maaten and Geoffrey Hinton, “Visualizing data using t-sne,” Journal of machine learning research, vol. 9, no. Nov, pp. 2579–2605, 2008

2008
[32]

Informa- tion theoretic measures for clusterings comparison: Variants, properties, normalization and correction for chance,

Nguyen Xuan Vinh, Julien Epps, and James Bailey, “Informa- tion theoretic measures for clusterings comparison: Variants, properties, normalization and correction for chance,” Journal of Machine Learning Research , vol. 11, no. Oct, pp. 2837– 2854, 2010

2010

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.

[1] [1]

Cross-lingual Text-independent Speaker Verification using Unsupervised Adversarial Discriminative Domain Adaptation

INTRODUCTION Speaker veriﬁcation (SV) offers a natural and ﬂexible option for bio- metric authentication. The text-independent SV system, which does not require the ﬁxed input voice content, is a ﬂexible and challenging task. In real-world scenarios, however, speaker veriﬁcation systems may degrade signiﬁcantly when training on one language and test it on...

work page Pith review arXiv 1908

[2] [2]

SPEAKER VERIFICATION SYSTEMS 2.1. The X-vector system We use a recently proposed successful speaker model called X- vector [6], to extract speaker representations, and a Probabilistic Linear Discriminant Analysis (PLDA) back-end to compare pairs of enrollment and test speaker embeddings. The X-vector model is based on a Time-Delay Deep Neural Network (TDN...

work page

[3] [3]

English Corpora We use Speaker Recognition Evaluation (SRE) 04-08, Mixer 6, and Switchboard (SWBD) to train the x-vector model

EXPERIMENTAL SETUP 3.1. English Corpora We use Speaker Recognition Evaluation (SRE) 04-08, Mixer 6, and Switchboard (SWBD) to train the x-vector model. SRE corpus is part of the Mixer 6 project, which was designed to support the devel- opment of robust speaker recognition technology by providing care- fully collected speech across numerous microphones. Sw...

work page

[4] [4]

0764”, it is difﬁcult to separate it from speaker “0765

RESULTS AND DISCUSSIONS 4.1. Results In this section, we show experimental results using x-vector, x- vector with DAT and x-vector with ADDA training with and without PLDA adaptations in Table 1. We use Linear Discriminant Analysis (LDA) to reduce all three embeddings to 256 dimension for compar- ison. Also, we concatenate the DAT embedding with the x-vec...

work page

[5] [5]

CONCLUSIONS AND FUTURE WORK We presented a discriminative adversarial unsupervised adaptation method in this paper. By exploiting how to alleviate the domain mismatch problem in an English-Chinese cross-lingual speaker ver- iﬁcation task, we showed that our proposed unsupervised ADDA approach can perform well on speaker classiﬁcation for the target domain...

work page

[6] [6]

Joint factor analysis versus eigenchannels in speaker recognition,

Patrick Kenny, Gilles Boulianne, Pierre Ouellet, and Pierre Dumouchel, “Joint factor analysis versus eigenchannels in speaker recognition,” IEEE Transactions on Audio, Speech, and Language Processing, vol. 15, no. 4, pp. 1435–1447, 2007

work page 2007

[7] [7]

Full-covariance ubm and heavy-tailed plda in i- vector speaker veriﬁcation,

Pavel Mat ˇejka, Ondˇrej Glembek, Fabio Castaldo, Md Jahangir Alam, Oldˇrich Plchot, Patrick Kenny, Luk ´aˇs Burget, and Jan ˇCernocky, “Full-covariance ubm and heavy-tailed plda in i- vector speaker veriﬁcation,” in Acoustics, Speech and Sig- nal Processing (ICASSP), IEEE International Conference on , 2011, pp. 4828–4831

work page 2011

[8] [8]

Speaker recognition by machines and humans: A tutorial review,

John HL Hansen and Tauﬁq Hasan, “Speaker recognition by machines and humans: A tutorial review,” IEEE Signal pro- cessing magazine, vol. 32, no. 6, pp. 74–99, 2015

work page 2015