Language Adaptation for Speaker Recognition Systems Using Contrastive Learning

Brignatz, Vincent; Duret, Jarod; Matrouf, Driss; Rouvier, Mickaël

doi:10.1007/978-3-030-87802-3_9

Cited by 3 publications

(4 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Language dependency can make the system less usable as users may belong to different geographic locations and speak varied languages. It is challenging to get labeled datasets for various low-resource languages (Brignatz et al, 2021). Moreover, when tested on multilingual datasets and features, the models that show consistent behavior may be helpful in other applications (such as code-switching) through information sharing (Belinkov et al, 2019).…”

Section: Background and Motivationmentioning

confidence: 99%

“…Recent works: Transfer learning is a solution to address the problem of domain mismatch. However, it is challenging to get labeled datasets for various low-resource languages (Brignatz et al, 2021). Recent works investigate adversarial domain adaptation techniques for solving cross-lingual speaker verification problems (Rohdin et al, 2019;Xia et al, 2019;Chen et al, 2020;Brignatz et al, 2021).…”

Section: Background and Motivationmentioning

confidence: 99%

“…However, it is challenging to get labeled datasets for various low-resource languages (Brignatz et al, 2021). Recent works investigate adversarial domain adaptation techniques for solving cross-lingual speaker verification problems (Rohdin et al, 2019;Xia et al, 2019;Chen et al, 2020;Brignatz et al, 2021). However, most of these approaches can improve the performance of speaker verification models for a limited set of languages as these approaches require an additional overhead of domain adaptation.…”

Section: Background and Motivationmentioning

confidence: 99%

See 2 more Smart Citations

FAtNet: Cost-Effective Approach Towards Mitigating the Linguistic Bias in Speaker Verification Systems

Sharma¹,

Buduru²

2022

Findings of the Association for Computational Linguistics: NAACL 2022

View full text Add to dashboard Cite

Linguistic bias in Deep Neural Network (DNN) based Natural Language Processing (NLP) systems is a critical problem that needs attention. The problem further intensifies in the case of security systems, such as speaker verification, where fairness is essential. Speaker verification systems are intelligent systems that determine if two speech recordings belong to the same speaker. Such human-oriented security systems should be usable by diverse people speaking varied languages. Thus, a speaker verification system trained on speech in one language should generalize when tested for other languages. However, DNN-based models are often language-dependent. Previous works explore domain adaptation to fine-tune the pre-trained model for out-of-domain languages. Fine-tuning the model individually for each existing language is expensive. Hence, it limits the usability of the system. This paper proposes the cost-effective idea of integrating a lightweight embedding with existing speaker verification systems to mitigate linguistic bias without adaptation. This work is motivated by the theoretical hypothesis that attentive-frames could help generate language-agnostic embeddings. For scientific validation of this hypothesis, we propose two frame-attentive networks and investigate the effect of their integration with baselines for twelve languages. Empirical results suggest that frame-attentive embedding can cost-effectively reduce linguistic bias and enhance the usability of baselines.

show abstract

Section: Background and Motivationmentioning

confidence: 99%

Section: Background and Motivationmentioning

confidence: 99%

Section: Background and Motivationmentioning

confidence: 99%

See 1 more Smart Citation

FAtNet: Cost-Effective Approach Towards Mitigating the Linguistic Bias in Speaker Verification Systems

Sharma¹,

Buduru²

2022

Findings of the Association for Computational Linguistics: NAACL 2022

View full text Add to dashboard Cite

show abstract

“…Several self-supervised learning methods are proposed for robust data representation [13] [14] . Among the proposed methods contrastive loss function is applied in robust speaker recognition system [3] and language adaptation [15]. Although the contrastive loss function has given promising results in domain adaption and robust speaker recognition, it has some limitations such as necessity for large batch size and the way of defining the negative pairs [16].…”

Section: Introductionmentioning

confidence: 99%

Barlow Twins self-supervised learning for robust speaker recognition

et al. 2022

Self Cite

View full text Add to dashboard Cite

Acoustic noise is a big challenge for speaker recognition systems. The state-of-the-art speaker recognition systems are based on deep neural network speaker embeddings called xvector extractor. A noise-robust x-vector extractor is highly demanded in speaker recognition systems. In this paper, we introduce Barlow Twins self-supervised loss function in the area of speaker recognition. Barlow Twins objective function tries to optimize two criteria: Firstly, it increases the similarity between two versions of the same signal (i.e. the clean and its augmented noisy version) to make the speaker embedding invariant to the acoustic noise. Secondly, it reduces the redundancy between dimensions of the x-vectors that improves the overall quality of speaker embeddings. In our research, Barlow Twins objective function is integrated with the ResNet-based speaker embedding system. In the proposed system, the Barlow Twins objective function is calculated in the embedding layer and it is optimized jointly with the speaker classifier loss function. The experimental results on Fabiole corpus show 22 % relative gain in terms of EER in the clean environments and 18% improvement in the presence of noise with low SNR and reverberation.

show abstract

A Comprehensive Exploration of Noise Robustness and Noise Compensation in ResNet and TDNN-based Speaker Recognition Systems

Mohammadamini

Matrouf

Bonastre

et al. 2022

2022 30th European Signal Processing Conference (EUSIPCO)

Self Cite

View full text Add to dashboard Cite

In this paper, a comprehensive exploration of noise robustness and noise compensation of ResNet and TDNN speaker recognition systems is presented. Firstly the robustness of the TDNN and ResNet in the presence of noise, reverberation, and both distortions is explored. Our experimental results show that in all cases the ResNet system is more robust than TDNN. After that, a noise compensation task is done with denoising autoencoder (DAE) over the x-vectors extracted from both systems. We explored two scenarios: 1) compensation of artificial noise with artificial data, 2) compensation of real noise with artificial data. The second case is the most desired scenario, because it makes noise compensation affordable without having real data to train denoising techniques. The experimental results show that in the first scenario noise compensation gives significant improvement with TDNN while this improvement in Resnet is not significant. In the second scenario, we achieved 15% improvement of EER over VoiCes Eval challenge in both TDNN and ResNet systems. In most cases the performance of ResNet without compensation is superior to TDNN with noise compensation.

show abstract

Language Adaptation for Speaker Recognition Systems Using Contrastive Learning

Cited by 3 publications

References 11 publications

FAtNet: Cost-Effective Approach Towards Mitigating the Linguistic Bias in Speaker Verification Systems

FAtNet: Cost-Effective Approach Towards Mitigating the Linguistic Bias in Speaker Verification Systems

Barlow Twins self-supervised learning for robust speaker recognition

A Comprehensive Exploration of Noise Robustness and Noise Compensation in ResNet and TDNN-based Speaker Recognition Systems

Contact Info

Product

Resources

About