Speech Enhancement Using End-to-End Speech Recognition Objectives

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F00216305%3A26230%2F19%3APU140038" target="_blank" >RIV/00216305:26230/19:PU140038 - isvavai.cz</a>

Výsledek na webu

<a href="https://ieeexplore.ieee.org/document/8937250" target="_blank" >https://ieeexplore.ieee.org/document/8937250</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.1109/WASPAA.2019.8937250" target="_blank" >10.1109/WASPAA.2019.8937250</a>

Jazyk výsledku

angličtina

Název v původním jazyce

Speech Enhancement Using End-to-End Speech Recognition Objectives

Popis výsledku v původním jazyce

Speech enhancement systems, which denoise and dereverberate distorted signals, are usually optimized based on signal reconstruction objectives including the maximum likelihood and minimum mean square error. However, emergent end-to-end neural methods enable to optimize the speech enhancement system with more applicationoriented objectives. For example, we can jointly optimize speech enhancement and automatic speech recognition (ASR) only with ASR error minimization criteria. The major contribution of this paper is to investigate how a system optimized based on the ASR objective improves the speech enhancement quality on various signal level metrics in addition to the ASR word error rate (WER) metric. We use a recently developed multichannel end-to-end (ME2E) system, which integrates neural dereverberation, beamforming, and attention-based speech recognition within a single neural network. Additionally, we propose to extend the dereverberation sub network of ME2E by dynamically varying the filter order in linear prediction by using reinforcement learning, and extend the beamforming subnetwork by incorporating the estimation of a speech distortion factor. The experiments reveal how well different signal level metrics correlate with the WER metric, and verify that learning-based speech enhancement can be realized by end-to-end ASR training objectives without using parallel clean and noisy data.

Název v anglickém jazyce

Speech Enhancement Using End-to-End Speech Recognition Objectives

Popis výsledku anglicky

Speech enhancement systems, which denoise and dereverberate distorted signals, are usually optimized based on signal reconstruction objectives including the maximum likelihood and minimum mean square error. However, emergent end-to-end neural methods enable to optimize the speech enhancement system with more applicationoriented objectives. For example, we can jointly optimize speech enhancement and automatic speech recognition (ASR) only with ASR error minimization criteria. The major contribution of this paper is to investigate how a system optimized based on the ASR objective improves the speech enhancement quality on various signal level metrics in addition to the ASR word error rate (WER) metric. We use a recently developed multichannel end-to-end (ME2E) system, which integrates neural dereverberation, beamforming, and attention-based speech recognition within a single neural network. Additionally, we propose to extend the dereverberation sub network of ME2E by dynamically varying the filter order in linear prediction by using reinforcement learning, and extend the beamforming subnetwork by incorporating the estimation of a speech distortion factor. The experiments reveal how well different signal level metrics correlate with the WER metric, and verify that learning-based speech enhancement can be realized by end-to-end ASR training objectives without using parallel clean and noisy data.

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/LQ1602" target="_blank" >LQ1602: IT4Innovations excellence in science</a><br>

Návaznosti

P - Projekt vyzkumu a vyvoje financovany z verejnych zdroju (s odkazem do CEP)<br>S - Specificky vyzkum na vysokych skolach

Rok uplatnění

2019

Kód důvěrnosti údajů

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

Název statě ve sborníku

IEEE Workshop on Applications of Signal Processing to Audio and Acoustics

ISBN

978-1-7281-1123-0

ISSN

—

e-ISSN

—

Počet stran výsledku

5

Strana od-do

234-238

Název nakladatele

IEEE Signal Processing Society

Místo vydání

New Paltz, NY

Místo konání akce

New Paltz, NY

Datum konání akce

20. 10. 2019

Typ akce podle státní příslušnosti

WRD - Celosvětová akce

Kód UT WoS článku

000527800200048