Predicting Hungarian sound durations for continuous speech Cover Image

Predicting Hungarian sound durations for continuous speech
Predicting Hungarian sound durations for continuous speech

Author(s): Gábor Olaszy
Subject(s): Phonetics / Phonology, Finno-Ugrian studies
Published by: Akadémiai Kiadó
Keywords: speech synthesis; speech synthesis; text-to-speech conversion; sound duration; speech unit inventory; suprasegmental duration modifications;

Summary/Abstract: Direct measurements show that a number of factors influence the final value of sound durations in continuous speech. On the segmental level it is mainly the articulatory movements that determine important influence factors, while on the suprasegmental level accent, syllabic stress, within-word position, the preceding and following syllables and finallyutterance position may have an influence on final sound durations. So the problem of how to predict sound durations can be described with a multivariable function in which the effect of the variables cannot be easily defined with good accuracy. It is difficult to separate the effects of certain functions, i.e., it is difficult to model this function, making direct measurements on the speech signal. A model has been constructed and realized in which three well-defined levels are working separately. In the first one (this is the segmental level) the separation of the effect of articulation from other factors is solved. The second and third levels relate to the suprasegmental level of speech.

  • Issue Year: 49/2002
  • Issue No: 3-4
  • Page Range: 321-345
  • Page Count: 25
  • Language: English