Robust Features for Speaker-Independent Speech Recognition Based on a Certain Class of Translation-Invariant Transformations

Florian Müller; Alfred Mertins

doi:10.1007/978-3-642-11509-7_15

Robust Features for Speaker-Independent Speech Recognition Based on a Certain Class of Translation-Invariant Transformations

Florian Müller, Alfred Mertins

Institute of Signal Processing

4 Citations (Scopus)

Abstract

The spectral effects of vocal tract length (VTL) differences are one reason for the lower recognition rate of today's speaker-independent automatic speech recognition (ASR) systems compared to speaker-dependent ones. By using certain types of filter banks the VTL-related effects can be described by a translation in subband-index space. In this paper, nonlinear translation-invariant transformations that originally have been proposed in the field of pattern recognition are investigated for their applicability in speaker-independent ASR tasks. It is shown that the combination of different types of such transformations leads to features that are more robust against VTL changes than the standard mel-frequency cepstral coefficients and that they almost yield the performance of vocal tract length normalization without any adaption to individual speakers.

Original language	English
Title of host publication	Advances in Nonlinear Speech Processing
Editors	Jordi Solé-Casals, Vladimir Zaiats
Number of pages	9
Volume	5933
Place of Publication	Berlin, Heidelberg
Publisher	Springer Berlin Heidelberg
Publication date	2010
Pages	111-119
ISBN (Print)	978-3-642-11508-0
ISBN (Electronic)	978-3-642-11509-7
DOIs	https://doi.org/10.1007/978-3-642-11509-7_15
Publication status	Published - 2010
Event	International Conference on Nonlinear Speech Processing 2009 - Vic, Spain Duration: 25.06.2009 → 27.06.2009 Conference number: 79990

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

SDG 9 Industry, Innovation, and Infrastructure

Access to Document

10.1007/978-3-642-11509-7_15

Cite this

Müller, F., & Mertins, A. (2010). Robust Features for Speaker-Independent Speech Recognition Based on a Certain Class of Translation-Invariant Transformations. In J. Solé-Casals, & V. Zaiats (Eds.), Advances in Nonlinear Speech Processing (Vol. 5933, pp. 111-119). ( Lecture Notes in Computer Science; Vol. 5933). Springer Berlin Heidelberg. https://doi.org/10.1007/978-3-642-11509-7_15

@inproceedings{6009e6ef35014b63af5a1a72e2474d1a,

title = "Robust Features for Speaker-Independent Speech Recognition Based on a Certain Class of Translation-Invariant Transformations",

abstract = "The spectral effects of vocal tract length (VTL) differences are one reason for the lower recognition rate of today's speaker-independent automatic speech recognition (ASR) systems compared to speaker-dependent ones. By using certain types of filter banks the VTL-related effects can be described by a translation in subband-index space. In this paper, nonlinear translation-invariant transformations that originally have been proposed in the field of pattern recognition are investigated for their applicability in speaker-independent ASR tasks. It is shown that the combination of different types of such transformations leads to features that are more robust against VTL changes than the standard mel-frequency cepstral coefficients and that they almost yield the performance of vocal tract length normalization without any adaption to individual speakers.",

author = "Florian M{\"u}ller and Alfred Mertins",

year = "2010",

doi = "10.1007/978-3-642-11509-7\_15",

language = "English",

isbn = "978-3-642-11508-0",

volume = "5933",

series = " Lecture Notes in Computer Science",

publisher = "Springer Berlin Heidelberg",

pages = "111--119",

editor = "Jordi Sol{\'e}-Casals and Vladimir Zaiats",

booktitle = "Advances in Nonlinear Speech Processing",

note = "International Conference on Nonlinear Speech Processing 2009, NOLISP 2009 ; Conference date: 25-06-2009 Through 27-06-2009",

}

Müller, F & Mertins, A 2010, Robust Features for Speaker-Independent Speech Recognition Based on a Certain Class of Translation-Invariant Transformations. in J Solé-Casals & V Zaiats (eds), Advances in Nonlinear Speech Processing. vol. 5933, Lecture Notes in Computer Science, vol. 5933, Springer Berlin Heidelberg, Berlin, Heidelberg, pp. 111-119, International Conference on Nonlinear Speech Processing 2009, Vic, Spain, 25.06.09. https://doi.org/10.1007/978-3-642-11509-7_15

Robust Features for Speaker-Independent Speech Recognition Based on a Certain Class of Translation-Invariant Transformations. / Müller, Florian; Mertins, Alfred.
Advances in Nonlinear Speech Processing. ed. / Jordi Solé-Casals; Vladimir Zaiats. Vol. 5933 Berlin, Heidelberg: Springer Berlin Heidelberg, 2010. p. 111-119 ( Lecture Notes in Computer Science; Vol. 5933).

Research output: Chapters in Books/Reports/Conference Proceedings › Conference contribution › peer-review

TY - GEN

T1 - Robust Features for Speaker-Independent Speech Recognition Based on a Certain Class of Translation-Invariant Transformations

AU - Müller, Florian

AU - Mertins, Alfred

N1 - Conference code: 79990

PY - 2010

Y1 - 2010

N2 - The spectral effects of vocal tract length (VTL) differences are one reason for the lower recognition rate of today's speaker-independent automatic speech recognition (ASR) systems compared to speaker-dependent ones. By using certain types of filter banks the VTL-related effects can be described by a translation in subband-index space. In this paper, nonlinear translation-invariant transformations that originally have been proposed in the field of pattern recognition are investigated for their applicability in speaker-independent ASR tasks. It is shown that the combination of different types of such transformations leads to features that are more robust against VTL changes than the standard mel-frequency cepstral coefficients and that they almost yield the performance of vocal tract length normalization without any adaption to individual speakers.

AB - The spectral effects of vocal tract length (VTL) differences are one reason for the lower recognition rate of today's speaker-independent automatic speech recognition (ASR) systems compared to speaker-dependent ones. By using certain types of filter banks the VTL-related effects can be described by a translation in subband-index space. In this paper, nonlinear translation-invariant transformations that originally have been proposed in the field of pattern recognition are investigated for their applicability in speaker-independent ASR tasks. It is shown that the combination of different types of such transformations leads to features that are more robust against VTL changes than the standard mel-frequency cepstral coefficients and that they almost yield the performance of vocal tract length normalization without any adaption to individual speakers.

UR - https://www.scopus.com/pages/publications/77951480608

U2 - 10.1007/978-3-642-11509-7_15

DO - 10.1007/978-3-642-11509-7_15

M3 - Conference contribution

SN - 978-3-642-11508-0

VL - 5933

T3 - Lecture Notes in Computer Science

SP - 111

EP - 119

BT - Advances in Nonlinear Speech Processing

A2 - Solé-Casals, Jordi

A2 - Zaiats, Vladimir

PB - Springer Berlin Heidelberg

CY - Berlin, Heidelberg

T2 - International Conference on Nonlinear Speech Processing 2009

Y2 - 25 June 2009 through 27 June 2009

ER -

Müller F, Mertins A. Robust Features for Speaker-Independent Speech Recognition Based on a Certain Class of Translation-Invariant Transformations. In Solé-Casals J, Zaiats V, editors, Advances in Nonlinear Speech Processing. Vol. 5933. Berlin, Heidelberg: Springer Berlin Heidelberg. 2010. p. 111-119. ( Lecture Notes in Computer Science). doi: 10.1007/978-3-642-11509-7_15

Robust Features for Speaker-Independent Speech Recognition Based on a Certain Class of Translation-Invariant Transformations

Abstract

UN SDGs

Access to Document

Other files and links

Fingerprint

Cite this