Pitch synchronous transform warping in voice conversion

0396886 - ÚFE 2014 RIV DE eng C - Konferenční příspěvek (zahraniční konf.)
Vích, Robert - Vondra, Martin
Pitch synchronous transform warping in voice conversion.
Lecture Notes i nComputer Science. Vol. 7403. Berlin: Springer Verlag, 2012, s. 280-289. ISBN 9783642345838. ISSN 0302-9743.
[International Training School on Cognitive Behavioural Systems, COST 2102. Dresden (DE), 21.02.2011-26.02.2011]
Grant CEP: GA MŠMT OC08010
Institucionální podpora: RVO:67985882
Klíčová slova: frequency transformation * speech analysis * speech synthesis
Kód oboru RIV: JA - Elektronika a optoelektronika, elektrotechnika

In this paper a new voice conversion algorithm is presented, which transforms the utterance of a source speaker into the utterance of a target speaker. The voice conversion approach is based on pitch synchronous speech analysis, Discrete Cosine Transform (DCT), nonlinear spectral warping with spectrum interpolation and pitch synchronous speech synthesis with overlapping using the speech production model. The DCT speech model contains also information about the phase properties of the modeled speech frame, but is, in contrary to a model based e.g. on the discrete Fourier transform, a real model and can be efficiently used for speech coding and voice conversion. The resulting finite impulse response of the converted DCT speech model is obtained by the inverse DCT and it is of the mixed phase type. The proposed voice conversion procedure results in speech with high naturalness
Trvalý link: http://hdl.handle.net/11104/0224614