Robust Automatic Recognition of Speech with Background Music

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F46747885%3A24220%2F17%3A00004811" target="_blank" >RIV/46747885:24220/17:00004811 - isvavai.cz</a>

Výsledek na webu

<a href="http://dx.doi.org/10.1109/ICASSP.2017.7953150" target="_blank" >http://dx.doi.org/10.1109/ICASSP.2017.7953150</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.1109/ICASSP.2017.7953150" target="_blank" >10.1109/ICASSP.2017.7953150</a>

Jazyk výsledku

angličtina

Název v původním jazyce

Robust Automatic Recognition of Speech with Background Music

Popis výsledku v původním jazyce

This paper addresses the task of Automatic Speech Recognition (ASR) with music in the background, where the accuracy of recognition may deteriorate significantly. To improve the robustness of ASR in this task, e.g. for broadcast news transcription or subtitles creation, we adopt two approaches: 1) multi-condition training of the acoustic models and 2) denoising autoencoders followed by acoustic model training on the preprocessed data. In the latter case, two types of autoencoders are considered: the fully connected and the convolutional network. Presented experimental results show that all the investigated techniques are able to improve the recognition of speech distorted by music significantly. For example, in the case of artificial mixtures of speech and electronic music (low Signal-to-Noise Ratio (SNR) of 0 dB), we achieved absolute improvement of accuracy by 35.8%. For real-world broadcast news and a high SNR (about 10 dB), we achieved improvement by 2.4%. The important advantage of the studied approaches is that they do not deteriorate the accuracy in scenarios with clean speech (the decrease is about 1%).

Název v anglickém jazyce

Robust Automatic Recognition of Speech with Background Music

Popis výsledku anglicky

This paper addresses the task of Automatic Speech Recognition (ASR) with music in the background, where the accuracy of recognition may deteriorate significantly. To improve the robustness of ASR in this task, e.g. for broadcast news transcription or subtitles creation, we adopt two approaches: 1) multi-condition training of the acoustic models and 2) denoising autoencoders followed by acoustic model training on the preprocessed data. In the latter case, two types of autoencoders are considered: the fully connected and the convolutional network. Presented experimental results show that all the investigated techniques are able to improve the recognition of speech distorted by music significantly. For example, in the case of artificial mixtures of speech and electronic music (low Signal-to-Noise Ratio (SNR) of 0 dB), we achieved absolute improvement of accuracy by 35.8%. For real-world broadcast news and a high SNR (about 10 dB), we achieved improvement by 2.4%. The important advantage of the studied approaches is that they do not deteriorate the accuracy in scenarios with clean speech (the decrease is about 1%).

Druh

D - Stať ve sborníku

CEP obor

—

OECD FORD obor

20204 - Robotics and automatic control

Projekt

<a href="/cs/project/TA04010199" target="_blank" >TA04010199: MULTILINMEDIA - Multilinguální platforma pro monitoring a analýzu multimédií</a><br>

Návaznosti

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

Rok uplatnění

2017

Kód důvěrnosti údajů

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

Název statě ve sborníku

16 June 2017, Article number 7953150, Pages 5210-52142017 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2017; Hilton New Orleans RiversideNew Orleans; United States; 5 March 2017 through 9 March 2017; Category numberCFP

ISBN

978-1-5090-4117-6

ISSN

1520-6149

e-ISSN

—

Počet stran výsledku

5

Strana od-do

5210-5214

Název nakladatele

Institute of Electrical and Electronics Engineers Inc.

Místo vydání

USA

Místo konání akce

New Orleans, USA

Datum konání akce

1. 1. 2017

Typ akce podle státní příslušnosti

WRD - Celosvětová akce

Kód UT WoS článku

000414286205074