Detektor řečové aktivity založený na statistických metodách

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F00216305%3A26220%2F05%3APU51360" target="_blank" >RIV/00216305:26220/05:PU51360 - isvavai.cz</a>

Výsledek na webu

—

DOI - Digital Object Identifier

—

Jazyk výsledku

angličtina

Název v původním jazyce

Voice Activity Detection based-on Statistic Models

Popis výsledku v původním jazyce

This paper deals with the method of voice activity detection. Rising requirements to applications with variable-rate speech coding also increases the role of voice activity detectors which become crucial for the efficient bandwidth reduction. The paper aims especially to two detector types; these are the detector based on short-time signal energy and statistical model-based voice activity detector. Decision-making rules of the statistical model-based detector are derived from the LRT (Likelihood Ratio TTest) by estimating unknown parameters using the ML (Maximum Likelihood) criterion. In addition there is an effective hang-over scheme based on HMM (Hidden Markov Model) principles in the detector. There were models for these detectors created in Matlabsimulation environment. Models were tested on voice signals with the white noise as well as real noises. The results were evaluated and compared for various types of noises and SNR (Signal to Noise Ratios) values.

Název v anglickém jazyce

Voice Activity Detection based-on Statistic Models

Popis výsledku anglicky

This paper deals with the method of voice activity detection. Rising requirements to applications with variable-rate speech coding also increases the role of voice activity detectors which become crucial for the efficient bandwidth reduction. The paper aims especially to two detector types; these are the detector based on short-time signal energy and statistical model-based voice activity detector. Decision-making rules of the statistical model-based detector are derived from the LRT (Likelihood Ratio TTest) by estimating unknown parameters using the ML (Maximum Likelihood) criterion. In addition there is an effective hang-over scheme based on HMM (Hidden Markov Model) principles in the detector. There were models for these detectors created in Matlabsimulation environment. Models were tested on voice signals with the white noise as well as real noises. The results were evaluated and compared for various types of noises and SNR (Signal to Noise Ratios) values.

Druh

D - Stať ve sborníku

CEP obor

JA - Elektronika a optoelektronika, elektrotechnika

OECD FORD obor

—

Projekt

Výsledek vznikl pri realizaci vícero projektů. Více informací v záložce Projekty.

Návaznosti

Z - Vyzkumny zamer (s odkazem do CEZ)

Rok uplatnění

2005

Kód důvěrnosti údajů

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

Název statě ve sborníku

Proceedings of the 16th Conference Electronic Speech Signal Processing

ISBN

3-938863-17-X

ISSN

—

e-ISSN

—

Počet stran výsledku

6

Strana od-do

175-180

Název nakladatele

Academy of Science of the Czech Republic, Institute of Radio Engineering and Electronics

Místo vydání

Prague, Czech Republic

Místo konání akce

Prague, Lichenstein Palace

Datum konání akce

26. 9. 2005

Typ akce podle státní příslušnosti

WRD - Celosvětová akce

Kód UT WoS článku

—