Publication detail

Autoregressive Modeling based Feature Extraction for Aurora3 DSR Task

MOTLÍČEK, P., ČERNOCKÝ, J.

Original Title

Autoregressive Modeling based Feature Extraction for Aurora3 DSR Task

Type

conference paper

Language

English

Original Abstract

Techniques for analysis of speech, that use autoregressive (all-pole) modeling approaches, are presented here and compared to generally known Mel-frequency cepstrum based feature extraction. In the paper, first, we focus on several possible applications of modeling speech power spectra that increase the performance of ASR system mainly in case of large mismatch between training and testing data. Then, the attention is payed to the different types of features that can be extracted from all-pole model to reduce the overall word error rate. The results show that generally used cepstrum based features, which can be easily extracted from all-pole model, are not the most suitable parameters for ASR, where the input speech is corrupted by different types of real noises. Very good recognition performances were achieved e.g., with discrete or selective all-pole modeling based approaches, or with decorrelated line spectral frequencies. The feature extraction techniques were tested on SpeechDat-Car databases used for front-end evaluation of advanced distributed speech recognition (DSR) systems.

Keywords

speech proceesing, speech recognition, all-pole model, PLP, MFCC, feature extraction

Authors

MOTLÍČEK, P., ČERNOCKÝ, J.

RIV year

2003

Released

17. 6. 2003

Publisher

Institute for Perceptual Artificial Intelligence

Location

Geneva

ISBN

1018-4074

Periodical

European Conference EUROSPEECH

Year of study

2003

Number

9

State

Swiss Confederation

Pages from

1801

Pages to

1804

Pages count

4

URL

http://www.symporg.ch/eurospeech/, http://www.fit.vutbr.cz/~motlicek/publi/2003/motlicek_01.pdf

BibTex

@inproceedings{BUT14177,
  author="Petr {Motlíček} and Jan {Černocký}",
  title="Autoregressive Modeling based Feature Extraction for Aurora3 DSR Task",
  booktitle="Proc. EUROSPEECH 2003",
  year="2003",
  journal="European Conference EUROSPEECH",
  volume="2003",
  number="9",
  pages="1801--1804",
  publisher="Institute for Perceptual Artificial Intelligence",
  address="Geneva",
  issn="1018-4074",
  url="http://www.symporg.ch/eurospeech/, http://www.fit.vutbr.cz/~motlicek/publi/2003/motlicek_01.pdf"
}