Detail publikace

Discriminative Training Techniques for Acoustic Language Identification

BURGET, L. MATĚJKA, P. ČERNOCKÝ, J.

Originální název

Discriminative Training Techniques for Acoustic Language Identification

Anglický název

Discriminative Training Techniques for Acoustic Language Identification

Jazyk

en

Originální abstrakt

This paper presents comparison of Maximum Likelihood (ML) and  discriminative Maximum Mutual Information (MMI) training  for acoustic modeling in language identification (LID). Clear  advantage of MMI over ML training is shown. The final error rate compares favorably to other results published on NIST 2003 data.

Anglický abstrakt

This paper presents comparison of Maximum Likelihood (ML) and  discriminative Maximum Mutual Information (MMI) training  for acoustic modeling in language identification (LID). Clear  advantage of MMI over ML training is shown. The final error rate compares favorably to other results published on NIST 2003 data.

Dokumenty

BibTex


@inproceedings{BUT22218,
  author="Lukáš {Burget} and Pavel {Matějka} and Jan {Černocký}",
  title="Discriminative Training Techniques for Acoustic Language Identification",
  annote="This paper presents comparison of Maximum Likelihood (ML)
and  discriminative Maximum Mutual Information (MMI) training
 for acoustic modeling in language identification (LID). Clear  advantage of MMI
over ML training is shown. The final error rate compares favorably to other
results published on NIST 2003 data.",
  booktitle="Proceedings of ICASSP 2006",
  chapter="22218",
  year="2006",
  month="august",
  pages="209--212",
  type="conference paper"
}