Investigation of Deep Neural Networks for Robust Recognition of Nonlinearly Distorted Speech

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F46747885%3A24220%2F14%3A%230002971" target="_blank" >RIV/46747885:24220/14:#0002971 - isvavai.cz</a>

Výsledek na webu

—

DOI - Digital Object Identifier

—

Jazyk výsledku

angličtina

Název v původním jazyce

Investigation of Deep Neural Networks for Robust Recognition of Nonlinearly Distorted Speech

Popis výsledku v původním jazyce

This paper studies the use of hybrid context-dependent Deep Neural Network Hidden Markov Model (DNN-HMM) architecture for robust recognition of speech affected by realworld nonlinear distortions. We consider two types of distortions; a) signals distortedthrough overgained microphone preamplifier in the analog domain and b) recordings exhibiting unnatural spectral sparseness, caused by excessive denoising or low-bit-rate compression. We compare the performance of DNN-HMM architecture with that of the conventional system, based on context-dependent Gaussian Mixture Model (GMM)- HMMs, which applies channel/speaker adaptation and/or feature compensation in the front-end via Histogram Equalization (HEQ). We show that DNN-HMM architecture achieves a significantly lower Word Error Rate (WER) on the considered distorted datasets and that the obtained relative WER reduction is higher than 60%. We also investigate the usefulness of the feature compensation via HEQ for a DNN-HMM system and show

Název v anglickém jazyce

Investigation of Deep Neural Networks for Robust Recognition of Nonlinearly Distorted Speech

Popis výsledku anglicky

This paper studies the use of hybrid context-dependent Deep Neural Network Hidden Markov Model (DNN-HMM) architecture for robust recognition of speech affected by realworld nonlinear distortions. We consider two types of distortions; a) signals distortedthrough overgained microphone preamplifier in the analog domain and b) recordings exhibiting unnatural spectral sparseness, caused by excessive denoising or low-bit-rate compression. We compare the performance of DNN-HMM architecture with that of the conventional system, based on context-dependent Gaussian Mixture Model (GMM)- HMMs, which applies channel/speaker adaptation and/or feature compensation in the front-end via Histogram Equalization (HEQ). We show that DNN-HMM architecture achieves a significantly lower Word Error Rate (WER) on the considered distorted datasets and that the obtained relative WER reduction is higher than 60%. We also investigate the usefulness of the feature compensation via HEQ for a DNN-HMM system and show

Druh

D - Stať ve sborníku

CEP obor

JC - Počítačový hardware a software

OECD FORD obor

—

Projekt

<a href="/cs/project/TA01011142" target="_blank" >TA01011142: Automatická transkripce a indexace přednášek</a><br>

Návaznosti

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

Rok uplatnění

2014

Kód důvěrnosti údajů

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

Název statě ve sborníku

Proceedings of the 15th Annual Conference of the International Speech Communication Association (INTERSPEECH 2014)

ISBN

—

ISSN

2308-457X

e-ISSN

—

Počet stran výsledku

5

Strana od-do

363-367

Název nakladatele

ISCA

Místo vydání

Singapure

Místo konání akce

Singapure

Datum konání akce

1. 1. 2014

Typ akce podle státní příslušnosti

WRD - Celosvětová akce

Kód UT WoS článku

—