Digital Model of Vocal Tract and Vocal Folds for Articulatory Speech Synthesizer,

Abstract

It has been known that, theoretically, the best method to generate artificial speech is direct modeling a human speech production system. Since articulatory synthesis models the human vocal organs as precise as possible, it is potentially the most desirable method to produce various words and languages. However, articulatory synthesis has been rarely realized, because its computational load is considerably high and the progress is too much complex. This paper proposes another type of an articulatory synthesizer and digital models for vocal tract and vocal folds producing Korean vowel and continuous utterance. Positions of articulators are extracted based on MRI data of Korean. KellyLochbaum digital filter is used to implement synthesizer based on the Mermelstein’s vocal tract model and two-mass vocal folds model. Various timbre, length, stress and intonation are tested, and their results are presented. Singing a song is finally realized based on these results.