Home > Research > Publications & Outputs > Discovering dynamical models of speech using ph...

Research

Associated organisational unit

Electronic data

  • Kirkham2024

    Final published version, 264 KB, PDF document

    Available under license: CC BY: Creative Commons Attribution 4.0 International License

Links

Text available via DOI:

View graph of relations

Discovering dynamical models of speech using physics-informed machine learning

Research output: Contribution in Book/Report/Proceedings - With ISBN/ISSN › Conference contribution/Paper

Published

Standard

Discovering dynamical models of speech using physics-informed machine learning. / Kirkham, Sam.
Proceedings of the 13th International Seminar on Speech Production. ed. / Cécile Fougeron; Pascal Perrier. Autrans, France: International Speech Communication Association, 2024. p. 177-180.

Research output: Contribution in Book/Report/Proceedings - With ISBN/ISSN › Conference contribution/Paper

Harvard

Kirkham, S 2024, Discovering dynamical models of speech using physics-informed machine learning. in C Fougeron & P Perrier (eds), Proceedings of the 13th International Seminar on Speech Production. International Speech Communication Association, Autrans, France, pp. 177-180, ISSP 2024 : 13th International Seminar on Speech Production, Autrans, France, 13/05/24. https://doi.org/10.21437/issp.2024-45

APA

Kirkham, S. (2024). Discovering dynamical models of speech using physics-informed machine learning. In C. Fougeron, & P. Perrier (Eds.), Proceedings of the 13th International Seminar on Speech Production (pp. 177-180). International Speech Communication Association. https://doi.org/10.21437/issp.2024-45

Vancouver

Kirkham S. Discovering dynamical models of speech using physics-informed machine learning. In Fougeron C, Perrier P, editors, Proceedings of the 13th International Seminar on Speech Production. Autrans, France: International Speech Communication Association. 2024. p. 177-180 doi: 10.21437/issp.2024-45

Author

Kirkham, Sam. / Discovering dynamical models of speech using physics-informed machine learning. Proceedings of the 13th International Seminar on Speech Production. editor / Cécile Fougeron ; Pascal Perrier. Autrans, France : International Speech Communication Association, 2024. pp. 177-180

Bibtex

@inproceedings{bf805178e347494682513af24f767b05,
title = "Discovering dynamical models of speech using physics-informed machine learning",
abstract = "Spoken language is characterised by a high-dimensional and highly variable set of physical movements that unfold over time. What are the fundamental dynamical principles that underlie this signal? In this study, we demonstrate the use of physics- informed machine learning (sparse symbolic regression) for discovering new dynamical models of speech articulation. We first demonstrate the model discovery procedure on simulated data and show that the algorithm is able to discover the original model with near-perfect accuracy, even when the data contain extensive variation in duration, initial conditions and tar- get positions, as well as in the presence of added noise. We then demonstrate a proof-of-concept applying the same technique to empirical data, which reveals a small set of candidate dynamical models with increasing levels of complexity and accuracy.",
author = "Sam Kirkham",
year = "2024",
month = may,
day = "13",
doi = "10.21437/issp.2024-45",
language = "English",
pages = "177--180",
editor = "C{\'e}cile Fougeron and Pascal Perrier",
booktitle = "Proceedings of the 13th International Seminar on Speech Production",
publisher = " International Speech Communication Association",
note = "ISSP 2024 : 13th International Seminar on Speech Production ; Conference date: 13-05-2024 Through 17-05-2024",
url = "https://issp24.inviteo.fr/",

}

RIS

TY - GEN

T1 - Discovering dynamical models of speech using physics-informed machine learning

AU - Kirkham, Sam

N1 - Conference code: 13th

PY - 2024/5/13

Y1 - 2024/5/13

N2 - Spoken language is characterised by a high-dimensional and highly variable set of physical movements that unfold over time. What are the fundamental dynamical principles that underlie this signal? In this study, we demonstrate the use of physics- informed machine learning (sparse symbolic regression) for discovering new dynamical models of speech articulation. We first demonstrate the model discovery procedure on simulated data and show that the algorithm is able to discover the original model with near-perfect accuracy, even when the data contain extensive variation in duration, initial conditions and tar- get positions, as well as in the presence of added noise. We then demonstrate a proof-of-concept applying the same technique to empirical data, which reveals a small set of candidate dynamical models with increasing levels of complexity and accuracy.

AB - Spoken language is characterised by a high-dimensional and highly variable set of physical movements that unfold over time. What are the fundamental dynamical principles that underlie this signal? In this study, we demonstrate the use of physics- informed machine learning (sparse symbolic regression) for discovering new dynamical models of speech articulation. We first demonstrate the model discovery procedure on simulated data and show that the algorithm is able to discover the original model with near-perfect accuracy, even when the data contain extensive variation in duration, initial conditions and tar- get positions, as well as in the presence of added noise. We then demonstrate a proof-of-concept applying the same technique to empirical data, which reveals a small set of candidate dynamical models with increasing levels of complexity and accuracy.

U2 - 10.21437/issp.2024-45

DO - 10.21437/issp.2024-45

M3 - Conference contribution/Paper

SP - 177

EP - 180

BT - Proceedings of the 13th International Seminar on Speech Production

A2 - Fougeron, Cécile

A2 - Perrier, Pascal

PB - International Speech Communication Association

CY - Autrans, France

T2 - ISSP 2024 : 13th International Seminar on Speech Production

Y2 - 13 May 2024 through 17 May 2024

ER -