Publication detail

Frame based phoneme classification using large margin and kernel methods

PFEIFER, V. BALÍK, M.

Original Title

Frame based phoneme classification using large margin and kernel methods

English Title

Frame based phoneme classification using large margin and kernel methods

Type

conference paper

Language

en

Original Abstract

This Article deals with the frame-based phoneme classifiers. A new efficient sequence-based training algorithm is proposed along with the sequence-based classification function. This method had shown to be an effective trainign tool compared with the standard frame-based training algorithm. The proposed algorithm had been further completed with the non-linear transform (kernel methods). The proposed algorithm was evaluated over the TIMIT speech corpus.

English abstract

This Article deals with the frame-based phoneme classifiers. A new efficient sequence-based training algorithm is proposed along with the sequence-based classification function. This method had shown to be an effective trainign tool compared with the standard frame-based training algorithm. The proposed algorithm had been further completed with the non-linear transform (kernel methods). The proposed algorithm was evaluated over the TIMIT speech corpus.

Keywords

classifier, features, kernel, optimalization, phoneme, prototypes, speech, tree

RIV year

2010

Released

20.08.2010

ISBN

978-963-88981-0-4

Book

Telecomunications and Signal Processing - TSP 2010

Pages from

1

Pages to

4

Pages count

4

BibTex


@inproceedings{BUT25510,
  author="Václav {Pfeifer} and Miroslav {Balík}",
  title="Frame based phoneme classification using large margin and kernel methods",
  annote="This Article deals with the frame-based phoneme classifiers. A new efficient sequence-based training algorithm is proposed along with the sequence-based classification function. This method had shown to be an effective trainign tool compared with the standard frame-based training algorithm. The proposed algorithm had been further completed with the non-linear transform (kernel methods). The proposed algorithm was evaluated over the TIMIT speech corpus.",
  booktitle="Telecomunications and Signal Processing - TSP 2010",
  chapter="25510",
  howpublished="electronic, physical medium",
  year="2010",
  month="august",
  pages="1--4",
  type="conference paper"
}