Acoustics vs. articulation in articulatory speech synthesis: One vocal tract target configuration has more than one sound

Authors: Eva Lasarcyk


Articulatory speech synthesis can be used for speech production research to gain insight into articulation patterns and their acoustic counterparts, the speech sounds. It can be used e.g. to conduct perception experiments that study the relationship between articulation and fine phonetic detail in the acoustic domain. In a case study, we focus on articulatory details in German vowels. Results indicate that the transcription of vowel quality changes depending on the acoustic settings used. The goal of this contribution is to illustrate the importance of these acoustic settings in articulatory synthesis, and to increase awareness regarding these settings. From a user’s perspective, the selection of a specific synthesis strategy entails certain acoustic settings. They define the details of how a geometric vocal tract target configuration is rendered into a vocal tract area function. Again applying certain acoustic settings, this area function is then used in an aerodynamic-acoustic simulation to produce the speech signal. Depending on the acoustic settings, one underlying vocal tract target configuration can in the end produce several different sounds

Year: 2010
In session: Speech Synthesis
Pages: 104 to 111