Enhancing Vocal Tract Length Normalization with Elastic Registration for Automatic Speech Recognition

Florian Müller; Alfred Mertins

Enhancing Vocal Tract Length Normalization with Elastic Registration for Automatic Speech Recognition

Institute of Signal Processing

Abstract

Vocal tract length normalization (VTLN) is commonly applied utterance-wise with a warping function that makes the assumption of a linear dependence between the vocal tract length and the location of the formants. In this work we propose a datadriven method for enhancing the performance of systems that already use standard VTLN. The method is based on elastic registration to estimate optimal non-parametric transformations to further reduce inter-speaker variabilities. Results show that the proposed method can increase the performance of monophone systems such that it reaches that of a triphone system.

Original language	English
Title of host publication	Proc. Interspeech-2012
Number of pages	4
Place of Publication	Portland, USA
Publication date	01.09.2012
Pages	1362-1365
ISBN (Print)	978-162276759-5
Publication status	Published - 01.09.2012
Event	13th Annual Conference of the International Speech Communication Association 2012 - Portland, United States Duration: 09.09.2012 → 13.09.2012 Conference number: 97207

UN SDGs

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

SDG 9 Industry, Innovation, and Infrastructure

Access to Document

http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.302.3497

Cite this

@inproceedings{fa54279d9a66432992bfe7927c8b3649,

title = "Enhancing Vocal Tract Length Normalization with Elastic Registration for Automatic Speech Recognition",

abstract = "Vocal tract length normalization (VTLN) is commonly applied utterance-wise with a warping function that makes the assumption of a linear dependence between the vocal tract length and the location of the formants. In this work we propose a datadriven method for enhancing the performance of systems that already use standard VTLN. The method is based on elastic registration to estimate optimal non-parametric transformations to further reduce inter-speaker variabilities. Results show that the proposed method can increase the performance of monophone systems such that it reaches that of a triphone system.",

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

year = "2012",

month = sep,

day = "1",

language = "English",

isbn = " 978-162276759-5",

pages = "1362--1365",

booktitle = "Proc. Interspeech-2012",

note = "13th Annual Conference of the International Speech Communication Association 2012, INTERSPEECH 2012 ; Conference date: 09-09-2012 Through 13-09-2012",

}

Müller, F & Mertins, A 2012, Enhancing Vocal Tract Length Normalization with Elastic Registration for Automatic Speech Recognition. in Proc. Interspeech-2012. Portland, USA, pp. 1362-1365, 13th Annual Conference of the International Speech Communication Association 2012, Portland, Oregon, United States, 09.09.12. <http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.302.3497>

TY - GEN

T1 - Enhancing Vocal Tract Length Normalization with Elastic Registration for Automatic Speech Recognition

AU - Müller, Florian

AU - Mertins, Alfred

N1 - Conference code: 97207

PY - 2012/9/1

Y1 - 2012/9/1

N2 - Vocal tract length normalization (VTLN) is commonly applied utterance-wise with a warping function that makes the assumption of a linear dependence between the vocal tract length and the location of the formants. In this work we propose a datadriven method for enhancing the performance of systems that already use standard VTLN. The method is based on elastic registration to estimate optimal non-parametric transformations to further reduce inter-speaker variabilities. Results show that the proposed method can increase the performance of monophone systems such that it reaches that of a triphone system.

AB - Vocal tract length normalization (VTLN) is commonly applied utterance-wise with a warping function that makes the assumption of a linear dependence between the vocal tract length and the location of the formants. In this work we propose a datadriven method for enhancing the performance of systems that already use standard VTLN. The method is based on elastic registration to estimate optimal non-parametric transformations to further reduce inter-speaker variabilities. Results show that the proposed method can increase the performance of monophone systems such that it reaches that of a triphone system.

UR - https://www.scopus.com/pages/publications/84878420000

M3 - Conference contribution

SN - 978-162276759-5

SP - 1362

EP - 1365

BT - Proc. Interspeech-2012

CY - Portland, USA

T2 - 13th Annual Conference of the International Speech Communication Association 2012

Y2 - 9 September 2012 through 13 September 2012

ER -

Enhancing Vocal Tract Length Normalization with Elastic Registration for Automatic Speech Recognition

Abstract

UN SDGs

Access to Document

Other files and links

Fingerprint

Invariant features for automatic speech recognition based on complex models of speech production and auditory perception

Cite this

Enhancing Vocal Tract Length Normalization with Elastic Registration for Automatic Speech Recognition

Abstract

UN SDGs

Access to Document

Other files and links

Fingerprint

Projects

Invariant features for automatic speech recognition based on complex models of speech production and auditory perception

Cite this