EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Contrastive Speaker Representation Learning with Hard Negative Sampling for Speaker Recognition

Changhwan Go, Young Han Lee, Taewoo Kim iD, Nam In Park iD, Chanjun Chun iD

Sensors · 2024

DOI: 10.3390/s24196213 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Sensors, 2024. Verfasst von Changhwan Go und 4 weiteren Autor:innen. Beteiligte Einrichtung: Department of Computer Engineering, Chosun University, Gwangju 61452, Republic of Korea. Hauptthemen laut MEDLINE: Speech. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 54 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: This work was partially supported by the Institute of Information & Communications 407 Technology Planning & Evaluation (IITP) grant funded by the Korean gover…, This work was partially supported by a research fund from Chosun University, 2023., Institute of Information & Communications Technology Planning & Evaluation (IITP).

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Speaker recognition is a technology that identifies the speaker in an input utterance by extracting speaker-distinguishable features from the speech signal. Speaker recognition is used for system security and authentication; therefore, it is crucial to extract unique features of the speaker to achieve high recognition rates. Representative methods for extracting these features include a classification approach, or utilizing contrastive learning to learn the speaker relationship between representations and then using embeddings extracted from a specific layer of the model. This paper introduces a framework for developing robust speaker recognition models through contrastive learning. This approach aims to minimize the similarity to hard negative samples—those that are genuine negatives, but have extremely similar features to the positives, leading to potential mistaken. Specifically, our proposed method trains the model by estimating hard negative samples within a mini-batch during contrastive learning, and then utilizes a cross-attention mechanism to determine speaker agreement for pairs of utterances. To demonstrate the effectiveness of our proposed method, we compared the performance of a deep learning model trained with a conventional loss function utilized in speaker recognition with that of a deep learning model trained using our proposed method, as measured by the equal error rate (EER), an objective performance metric. Our results indicate that when trained with the voxceleb2 dataset, the proposed method achieved an EER of 0.98% on the voxceleb1-E dataset and 1.84% on the voxceleb1-H dataset.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

SpeechHumansAlgorithmsDeep LearningPattern Recognition, AutomatedSpeech Recognition Software

Stichwörter der Autor:innen

contrastive learninghard negative samplingspeaker recognition

Autor:innen und Institutionen

  • Department of Computer Engineering, Chosun University, Gwangju 61452, Republic of Korea.
  • Intelligent Image Processing Research Center, Korea Electronics Technology Institute, Seongnam 13509, Republic of Korea.
  • Digital Analysis Division, National Forensic Service, Wonju 26460, Republic of Korea.

ORCID-Profile: Taewoo Kim, Nam In Park, Chanjun Chun

Förderung

  • This work was partially supported by the Institute of Information & Communications 407 Technology Planning & Evaluation (IITP) grant funded by the Korean gover…
  • This work was partially supported by a research fund from Chosun University, 2023.
  • Institute of Information & Communications Technology Planning & Evaluation (IITP)
  • Korean government (MSIT)
  • Localization Technology Development on Spoken Language Synthesis and Translation of OTT Media Content
  • Chosun University

Zitieren

APA

Go, C., Lee, Y. H., Kim, T., Park, N. I., & Chun, C. (2024). Contrastive Speaker Representation Learning with Hard Negative Sampling for Speaker Recognition. Sensors. https://doi.org/10.3390/s24196213

Vancouver

Go C, Lee YH, Kim T, Park NI, Chun C. Contrastive Speaker Representation Learning with Hard Negative Sampling for Speaker Recognition. Sensors. 2024. doi:10.3390/s24196213

BibTeX
@article{Go2024,
  author = {Go, Changhwan and Lee, Young Han and Kim, Taewoo and Park, Nam In and Chun, Chanjun},
  title = {Contrastive Speaker Representation Learning with Hard Negative Sampling for Speaker Recognition},
  journal = {Sensors},
  year = {2024},
  doi = {10.3390/s24196213}
}
Kostenlos starten →