Cross-lingual speaker verification with domain-balanced hard prototype mining and language-dependent score normalization
- Author
- Jenthe Thienpondt (UGent) , Brecht Desplanques (UGent) and Kris Demuynck (UGent)
- Organization
- Abstract
- In this paper we describe the top-scoring IDLab submission for the text-independent task of the Short-duration Speaker Verification (SdSV) Challenge 2020. The main difficulty of the challenge exists in the large degree of varying phonetic overlap between the potentially cross-lingual trials, along with the limited availability of in-domain DeepMine Farsi training data. We introduce domain-balanced hard prototype mining to finetune the state-of-the-art ECAPA-TDNN x-vector based speaker embedding extractor. The sample mining technique efficiently exploits speaker distances between the speaker prototypes of the popular AAM-softmax loss function to construct challenging training batches that are balanced on the domain-level. To enhance the scoring of cross-lingual trials, we propose a language-dependent s-norm score normalization. The imposter cohort only contains data from the Farsi target-domain which simulates the enrollment data always being Farsi. In case a Gaussian-Backend language model detects the test speaker embedding to contain English, a cross-language compensation offset determined on the AAM-softmax speaker prototypes is subtracted from the maximum expected imposter mean score. A fusion of five systems with minor topological tweaks resulted in a final MinDCF and EER of 0.065 and 1.45% respectively on the SdSVC evaluation set.
Downloads
-
DS369.pdf
- full text (Published version)
- |
- open access
- |
- |
- 263.59 KB
Citation
Please use this url to cite or link to this publication: http://hdl.handle.net/1854/LU-8680075
- MLA
- Thienpondt, Jenthe, et al. “Cross-Lingual Speaker Verification with Domain-Balanced Hard Prototype Mining and Language-Dependent Score Normalization.” Proc. Interspeech 2020, International Speech Communication Association (ISCA), 2020, pp. 756–60, doi:10.21437/Interspeech.2020-2662.
- APA
- Thienpondt, J., Desplanques, B., & Demuynck, K. (2020). Cross-lingual speaker verification with domain-balanced hard prototype mining and language-dependent score normalization. In Proc. Interspeech 2020 (pp. 756–760). Online: International Speech Communication Association (ISCA). https://doi.org/10.21437/Interspeech.2020-2662
- Chicago author-date
- Thienpondt, Jenthe, Brecht Desplanques, and Kris Demuynck. 2020. “Cross-Lingual Speaker Verification with Domain-Balanced Hard Prototype Mining and Language-Dependent Score Normalization.” In Proc. Interspeech 2020, 756–60. International Speech Communication Association (ISCA). https://doi.org/10.21437/Interspeech.2020-2662.
- Chicago author-date (all authors)
- Thienpondt, Jenthe, Brecht Desplanques, and Kris Demuynck. 2020. “Cross-Lingual Speaker Verification with Domain-Balanced Hard Prototype Mining and Language-Dependent Score Normalization.” In Proc. Interspeech 2020, 756–760. International Speech Communication Association (ISCA). doi:10.21437/Interspeech.2020-2662.
- Vancouver
- 1.Thienpondt J, Desplanques B, Demuynck K. Cross-lingual speaker verification with domain-balanced hard prototype mining and language-dependent score normalization. In: Proc Interspeech 2020. International Speech Communication Association (ISCA); 2020. p. 756–60.
- IEEE
- [1]J. Thienpondt, B. Desplanques, and K. Demuynck, “Cross-lingual speaker verification with domain-balanced hard prototype mining and language-dependent score normalization,” in Proc. Interspeech 2020, Online, 2020, pp. 756–760.
@inproceedings{8680075,
abstract = {{In this paper we describe the top-scoring IDLab submission for the text-independent task of the Short-duration Speaker Verification (SdSV) Challenge 2020. The main difficulty of the challenge exists in the large degree of varying phonetic overlap between the potentially cross-lingual trials, along with the limited availability of in-domain DeepMine Farsi training data. We introduce domain-balanced hard prototype mining to finetune the state-of-the-art ECAPA-TDNN x-vector based speaker embedding extractor. The sample mining technique efficiently exploits speaker distances between the speaker prototypes of the popular AAM-softmax loss function to construct challenging training batches that are balanced on the domain-level. To enhance the scoring of cross-lingual trials, we propose a language-dependent s-norm score normalization. The imposter cohort only contains data from the Farsi target-domain which simulates the enrollment data always being Farsi. In case a Gaussian-Backend language model detects the test speaker embedding to contain English, a cross-language compensation offset determined on the AAM-softmax speaker prototypes is subtracted from the maximum expected imposter mean score. A fusion of five systems with minor topological tweaks resulted in a final MinDCF and EER of 0.065 and 1.45% respectively on the SdSVC evaluation set.}},
author = {{Thienpondt, Jenthe and Desplanques, Brecht and Demuynck, Kris}},
booktitle = {{Proc. Interspeech 2020}},
issn = {{1990-9772}},
language = {{eng}},
location = {{Online}},
pages = {{756--760}},
publisher = {{International Speech Communication Association (ISCA)}},
title = {{Cross-lingual speaker verification with domain-balanced hard prototype mining and language-dependent score normalization}},
url = {{http://dx.doi.org/10.21437/Interspeech.2020-2662}},
year = {{2020}},
}
- Altmetric
- View in Altmetric