Control of Klatt speech synthesizer with high-level parameters.
Corine A. Bickley, Kenneth Noble Stevens, David R. Williams · The Journal of the Acoustical Society of America · 1992
This paper describes a system for synthesis of speech with a terminal analog synthesizer like the Klatt synthesizer. The control of the synthesizer is accomplished with a small number (about 10) of high-level (HL) parameters that capture directly certain attributes of the state of the vocal tract during speech production. The multiplicity of low-level parameters in the synthesizer are derived from the HL parameters through a set of mapping relations. The HL parameters consist of (1) the fundamental frequency; (2) a specification of the vocal tract in terms of the first three or four natural frequencies of the tract assuming no opening to the glottis or the nose; (3) the areas of three major orifices in the vocal tract: the glottal opening, the velopharyngeal opening, and (for consonants) the principal constriction in the vocal tract; (4) a measure of active expansion or contraction of the vocal-tract volume for obstruents; (5) a measure of the efficiency of turbulence noise generation for obstruents. The mapping relations for obstruents involve the calculation of airflows, pressures, wall displacements, glottal waveform characteristics, vocal-tract losses, and noise due to turbulence. Some examples of the mapping relations for both obstruents and sonorants are given. [Research supported by the National Institutes of Health.]