Czech time-domain TTS system with sample-by-sample harmonically pitch-normalized speech segment database

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F49777513%3A23520%2F02%3A00076269" target="_blank" >RIV/49777513:23520/02:00076269 - isvavai.cz</a>

Výsledek na webu

—

DOI - Digital Object Identifier

—

Jazyk výsledku

angličtina

Název v původním jazyce

Czech time-domain TTS system with sample-by-sample harmonically pitch-normalized speech segment database

Popis výsledku v původním jazyce

Monotonously recorded speech corpus is required to achieve high segmental quality in TTS systems. We record our own speech corpora with professional speakers. But, for the speaker, it is usually not easy to satisfy the requirement of monotonicity. The easier and cheaper way to obtain speech corpus for TTS system would be to use some of publicly available speech records or speech corpora available on the market. But those cannot be expected to be recorded monotonously. This paper proposes our effort to cope with it.We try techniques similar to "spectral reharmonization". The off-line algorithm is applied pitch-synchronously on every segment in the speech segment database. We use the FFT algorithm to obtain a set of harmonic parameters for every sub-segment defined by the time instants of neighboring pitch-marks. Described pitch-normalization algorithm is performed on voiced parts of the segment only.

Název v anglickém jazyce

Czech time-domain TTS system with sample-by-sample harmonically pitch-normalized speech segment database

Popis výsledku anglicky

Monotonously recorded speech corpus is required to achieve high segmental quality in TTS systems. We record our own speech corpora with professional speakers. But, for the speaker, it is usually not easy to satisfy the requirement of monotonicity. The easier and cheaper way to obtain speech corpus for TTS system would be to use some of publicly available speech records or speech corpora available on the market. But those cannot be expected to be recorded monotonously. This paper proposes our effort to cope with it.We try techniques similar to "spectral reharmonization". The off-line algorithm is applied pitch-synchronously on every segment in the speech segment database. We use the FFT algorithm to obtain a set of harmonic parameters for every sub-segment defined by the time instants of neighboring pitch-marks. Described pitch-normalization algorithm is performed on voiced parts of the segment only.

Druh

D - Stať ve sborníku

CEP obor

JD - Využití počítačů, robotika a její aplikace

OECD FORD obor

—

Projekt

<a href="/cs/project/GA102%2F02%2F0124" target="_blank" >GA102/02/0124: Hlasové technologie v podpoře informační společnosti</a><br>

Návaznosti

Z - Vyzkumny zamer (s odkazem do CEZ)

Rok uplatnění

2002

Kód důvěrnosti údajů

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

Název statě ve sborníku

Czech time-domain TTS system with sample-by-sample harmonically pitch-normalized speech segment database

ISBN

8086269094

ISSN

—

e-ISSN

—

Počet stran výsledku

3

Strana od-do

44

Název nakladatele

Academy of Sciences of Czech Republic

Místo vydání

Prague

Místo konání akce

Prague

Datum konání akce

1. 1. 2002

Typ akce podle státní příslušnosti

CST - Celostátní akce

Kód UT WoS článku

—