TWiki> CSTR Web>HcExternal (05 Mar 2019, Main.cvbotinh)EditAttach

Entries for the Hurricane Challenge 2013

  natural 'Plain' TTS TTSLGP-DRC
clean
ssn
cs
Voices

TTS: baseline Text-To-Speech trained with 'Plain' speech

TTSLGP-DRC: TTS voice with adaptation to Lombard excitation and duration models (L) + glimpse-based spectral shapping (GP) + dynamic range compression (DRC)

Maskers

ssn - speech-shaped noise (Mid SNR=-4dB)

cs - female competing speaker (Mid SNR=-14dB)

Intermediate voices (clean samples only)

  TTS TTSL spec TTSL exc TTSL dur TTSL exc+dur TTSLomb
unmodified
GP
GP + DRC          
Voices

TTS: baseline Text-To-Speech trained with 'Plain' speech

TTSL spec: adaptation to Lombard spectral model

TTSL exc: adaptation to Lombard excitation model

TTSL dur: adaptation to Lombard duration model

TTSL exc+dur: adaptation to Lombard excitation and duration models

TTSLomb: adaptation to Lombard spec+exc+dur models

Modification

GP: noise-dependent Glimpse-based spectral modification to generated Mel cepstral coefficients

DRC: dynamic range compression

Topic revision: r6 - 05 Mar 2019 - 11:33:22 - Main.cvbotinh
 
This site is powered by the TWiki collaboration platformCopyright © by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback
This Wiki uses Cookies