Improving Machine Hearing on Limited Data Sets

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F00216305%3A26220%2F19%3APU133045" target="_blank" >RIV/00216305:26220/19:PU133045 - isvavai.cz</a>

Výsledek na webu

<a href="https://ieeexplore.ieee.org/document/8970740" target="_blank" >https://ieeexplore.ieee.org/document/8970740</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.1109/ICUMT48472.2019.8970740" target="_blank" >10.1109/ICUMT48472.2019.8970740</a>

Jazyk výsledku

angličtina

Název v původním jazyce

Improving Machine Hearing on Limited Data Sets

Popis výsledku v původním jazyce

Convolutional neural network (CNN) architectures have originated and revolutionized machine learning for images. In order to take advantage of CNNs in predictive modeling with audio data, standard FFT-based signal processing methods are often applied to convert the raw audio waveforms into an image-like representations (e.g. spectrograms). Even though conventional images and spectrograms differ in their feature properties, this kind of pre-processing reduces the amount of training data necessary for successful training. In this contribution we investigate how input and target representations interplay with the amount of available training data in a music information retrieval setting. We compare the standard mel-spectrogram inputs with a newly proposed representation, called Mel scattering. Furthermore, we investigate the impact of additional target data representations by using and augmented target loss function which incorporates unused available information. We observe that all proposed methods outperform the standard mel-transform representation when using a limited data set and discuss their strengths and limitations. The source code for reproducibility of our experiments as well as intermediate results and model checkpoints are available in an online repository.

Název v anglickém jazyce

Improving Machine Hearing on Limited Data Sets

Popis výsledku anglicky

Convolutional neural network (CNN) architectures have originated and revolutionized machine learning for images. In order to take advantage of CNNs in predictive modeling with audio data, standard FFT-based signal processing methods are often applied to convert the raw audio waveforms into an image-like representations (e.g. spectrograms). Even though conventional images and spectrograms differ in their feature properties, this kind of pre-processing reduces the amount of training data necessary for successful training. In this contribution we investigate how input and target representations interplay with the amount of available training data in a music information retrieval setting. We compare the standard mel-spectrogram inputs with a newly proposed representation, called Mel scattering. Furthermore, we investigate the impact of additional target data representations by using and augmented target loss function which incorporates unused available information. We observe that all proposed methods outperform the standard mel-transform representation when using a limited data set and discuss their strengths and limitations. The source code for reproducibility of our experiments as well as intermediate results and model checkpoints are available in an online repository.

Druh

D - Stať ve sborníku

CEP obor

—

OECD FORD obor

10201 - Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)

Projekt

<a href="/cs/project/LO1401" target="_blank" >LO1401: Interdisciplinární výzkum bezdrátových technologií</a><br>

Návaznosti

P - Projekt vyzkumu a vyvoje financovany z verejnych zdroju (s odkazem do CEP)

Rok uplatnění

2019

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Název statě ve sborníku

2019 The 11th International Congress on Ultra Modern Telecommunications and Control Systems (ICUMT)

ISBN

978-1-7281-5764-1

ISSN

—

e-ISSN

—

Počet stran výsledku

6

Strana od-do

1-6

Název nakladatele

IEEE

Místo vydání

Dublin, Ireland

Místo konání akce

Dublin

Datum konání akce

28. 10. 2019

Typ akce podle státní příslušnosti

WRD - Celosvětová akce

Kód UT WoS článku

000540651700016