Publication detail

Nonlinear adaptive models for speech enhancement algorithms

KOULA, I., ZEZULA, R.

Original Title

Nonlinear adaptive models for speech enhancement algorithms

Type

journal article - other

Language

English

Original Abstract

This paper is focused on comparison of the effectiveness of three different one channel speech enhancement algorithms. These speech enhancement algorithms are based on spectral subtraction method and on different approaches to noise spectrum estimation. The first algorithm estimates noise spectrum on the basis of its statistical characters. Next two algorithms estimate noise spectrum by nonlinear adaptive models. These models are arti-ficial neural networks and adaptive neural-fuzzy interface system. An algo-rithm giving efficiency enhancement of the obtained results was based on the hit rate recognition from the output of a phoneme based speech recognizer based on hidden Markov's models and implemented through the HTK toolkit.

Keywords

estimate of noise power spectral density, one-channel noise cancelation algorithm, spectral subtraction, artificial neural network, adaptive neural-fuzzy interface system

Authors

KOULA, I., ZEZULA, R.

Released

30. 9. 2007

Publisher

GESTS

Location

Korea

ISBN

1738-9682

Periodical

International Transactions on Communication and Signal Processing

Year of study

10

Number

7

State

Republic of Korea

Pages from

138

Pages to

145

Pages count

9

BibTex

@article{BUT44688,
  author="Ivan {Koula} and Radek {Zezula}",
  title="Nonlinear adaptive models for speech enhancement algorithms",
  journal="International Transactions on Communication and Signal Processing",
  year="2007",
  volume="10",
  number="7",
  pages="138--145",
  issn="1738-9682"
}