Neural Grey-Box Guitar Amplifier Modelling With Limited Data

Stepan Miklanek; Alec Wright; Vesa Välimäki; Jiri Schimmel

Neural Grey-Box Guitar Amplifier Modelling With Limited Data

Stepan Miklanek^*
, Alec Wright
, Vesa Välimäki
, Jiri Schimmel

^*Corresponding author for this work

Brno University of Technology

Research output: Chapter in Book/Report/Conference proceeding › Conference article in proceedings › Scientific › peer-review

8 Citations (Scopus)

288 Downloads (Pure)

Abstract

This paper combines recurrent neural networks (RNNs) with the discretised Kirchhoff nodal analysis (DK-method) to create a grey-box guitar amplifier model. Both the objective and subjective results suggest that the proposed model is able to outperform a baseline black-box RNN model in the task of modelling a guitar
amplifier, including realistically recreating the behaviour of the amplifier equaliser circuit, whilst requiring significantly less training data. Furthermore, we adapt the linear part of the DK-method in a deep learning scenario to derive multiple state-space filters simultaneously. We frequency sample the filter transfer functions in parallel and perform frequency domain filtering to considerably reduce the required training times compared to recursive state-space filtering. This study shows that it is a powerful idea to separately model the linear and nonlinear parts of a guitar amplifier using supervised learning.

Original language	English
Title of host publication	Proceedings of the 26th International Conference on Digital Audio Effects (DAFx23)
Editors	Federico Fontana, Silvin Willemsen
Place of Publication	Copenhagen, Denmark
Publisher	Aalborg University
Pages	151-158
Number of pages	8
Publication status	Published - 4 Sept 2023
MoE publication type	A4 Conference publication
Event	International Conference on Digital Audio Effects - Aalborg University Copenhagen, Copenhagen, Denmark Duration: 4 Sept 2023 → 7 Sept 2023 Conference number: 26 https://dafx23.create.aau.dk/

Publication series

Name	Proceedings of the International Conference on Digital Audio Effects
ISSN (Electronic)	2413-6689

Conference

Conference	International Conference on Digital Audio Effects
Abbreviated title	DAFx
Country/Territory	Denmark
City	Copenhagen
Period	04/09/2023 → 07/09/2023
Internet address	https://dafx23.create.aau.dk/

UN SDGs

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

SDG 9 Industry, Innovation, and Infrastructure

Keywords

Audio signal processing
deep learning
digital filter design

Access to Document

Miklanek2023-Neural-grey-box-guitar-amp
This is an open-access article distributed under the terms of the Creative Commons Attribution 4.0 International License, which permits unrestricted use, distribution, adaptation, and reproduction in any medium, provided the original author and source are credited.
Final published version, 558 KBLicence: CC BY

Cite this

Miklanek, Stepan ; Wright, Alec ; Välimäki, Vesa et al. / Neural Grey-Box Guitar Amplifier Modelling With Limited Data. Proceedings of the 26th International Conference on Digital Audio Effects (DAFx23). editor / Federico Fontana ; Silvin Willemsen. Copenhagen, Denmark : Aalborg University, 2023. pp. 151-158 (Proceedings of the International Conference on Digital Audio Effects).

@inproceedings{8661cb32cf13482497ab26b6f0742f94,

title = "Neural Grey-Box Guitar Amplifier Modelling With Limited Data",

abstract = "This paper combines recurrent neural networks (RNNs) with the discretised Kirchhoff nodal analysis (DK-method) to create a grey-box guitar amplifier model. Both the objective and subjective results suggest that the proposed model is able to outperform a baseline black-box RNN model in the task of modelling a guitar amplifier, including realistically recreating the behaviour of the amplifier equaliser circuit, whilst requiring significantly less training data. Furthermore, we adapt the linear part of the DK-method in a deep learning scenario to derive multiple state-space filters simultaneously. We frequency sample the filter transfer functions in parallel and perform frequency domain filtering to considerably reduce the required training times compared to recursive state-space filtering. This study shows that it is a powerful idea to separately model the linear and nonlinear parts of a guitar amplifier using supervised learning.",

keywords = "Audio signal processing, deep learning, digital filter design",

author = "Stepan Miklanek and Alec Wright and Vesa V{\"a}lim{\"a}ki and Jiri Schimmel",

year = "2023",

month = sep,

day = "4",

language = "English",

series = "Proceedings of the International Conference on Digital Audio Effects",

publisher = "Aalborg University",

pages = "151--158",

editor = "Federico Fontana and Silvin Willemsen",

booktitle = "Proceedings of the 26th International Conference on Digital Audio Effects (DAFx23)",

address = "Denmark",

note = "International Conference on Digital Audio Effects, DAFx ; Conference date: 04-09-2023 Through 07-09-2023",

url = "https://dafx23.create.aau.dk/",

}

Miklanek, S, Wright, A, Välimäki, V & Schimmel, J 2023, Neural Grey-Box Guitar Amplifier Modelling With Limited Data. in F Fontana & S Willemsen (eds), Proceedings of the 26th International Conference on Digital Audio Effects (DAFx23). Proceedings of the International Conference on Digital Audio Effects, Aalborg University, Copenhagen, Denmark, pp. 151-158, International Conference on Digital Audio Effects, Copenhagen, Denmark, 04/09/2023.

Neural Grey-Box Guitar Amplifier Modelling With Limited Data. / Miklanek, Stepan; Wright, Alec; Välimäki, Vesa et al.
Proceedings of the 26th International Conference on Digital Audio Effects (DAFx23). ed. / Federico Fontana; Silvin Willemsen. Copenhagen, Denmark: Aalborg University, 2023. p. 151-158 (Proceedings of the International Conference on Digital Audio Effects).

Research output: Chapter in Book/Report/Conference proceeding › Conference article in proceedings › Scientific › peer-review

TY - GEN

T1 - Neural Grey-Box Guitar Amplifier Modelling With Limited Data

AU - Miklanek, Stepan

AU - Wright, Alec

AU - Välimäki, Vesa

AU - Schimmel, Jiri

N1 - Conference code: 26

PY - 2023/9/4

Y1 - 2023/9/4

N2 - This paper combines recurrent neural networks (RNNs) with the discretised Kirchhoff nodal analysis (DK-method) to create a grey-box guitar amplifier model. Both the objective and subjective results suggest that the proposed model is able to outperform a baseline black-box RNN model in the task of modelling a guitar amplifier, including realistically recreating the behaviour of the amplifier equaliser circuit, whilst requiring significantly less training data. Furthermore, we adapt the linear part of the DK-method in a deep learning scenario to derive multiple state-space filters simultaneously. We frequency sample the filter transfer functions in parallel and perform frequency domain filtering to considerably reduce the required training times compared to recursive state-space filtering. This study shows that it is a powerful idea to separately model the linear and nonlinear parts of a guitar amplifier using supervised learning.

AB - This paper combines recurrent neural networks (RNNs) with the discretised Kirchhoff nodal analysis (DK-method) to create a grey-box guitar amplifier model. Both the objective and subjective results suggest that the proposed model is able to outperform a baseline black-box RNN model in the task of modelling a guitar amplifier, including realistically recreating the behaviour of the amplifier equaliser circuit, whilst requiring significantly less training data. Furthermore, we adapt the linear part of the DK-method in a deep learning scenario to derive multiple state-space filters simultaneously. We frequency sample the filter transfer functions in parallel and perform frequency domain filtering to considerably reduce the required training times compared to recursive state-space filtering. This study shows that it is a powerful idea to separately model the linear and nonlinear parts of a guitar amplifier using supervised learning.

KW - Audio signal processing

KW - deep learning

KW - digital filter design

UR - https://dafx23.create.aau.dk/wp-content/uploads/2023/09/DAFX23_Proceedings.pdf

M3 - Conference article in proceedings

T3 - Proceedings of the International Conference on Digital Audio Effects

SP - 151

EP - 158

BT - Proceedings of the 26th International Conference on Digital Audio Effects (DAFx23)

A2 - Fontana, Federico

A2 - Willemsen, Silvin

PB - Aalborg University

CY - Copenhagen, Denmark

T2 - International Conference on Digital Audio Effects

Y2 - 4 September 2023 through 7 September 2023

ER -

Neural Grey-Box Guitar Amplifier Modelling With Limited Data

Abstract

Publication series

Conference

UN SDGs

Keywords

Access to Document

Other files and links

Fingerprint

NordicSMC: Nordic Sound and Music Computing Network

Aalto Acoustics Lab

Štěpán Miklánek

Cite this

Neural Grey-Box Guitar Amplifier Modelling With Limited Data

Abstract

Publication series

Conference

UN SDGs

Keywords

Access to Document

Other files and links

Fingerprint

Projects

NordicSMC: Nordic Sound and Music Computing Network

Equipment

Aalto Acoustics Lab

Activities

Štěpán Miklánek

Cite this