Observations on the dynamic control of an articulatory synthesizer using speech production data

Ingmar Steiner

PhD-Thesis Saarland University SciDok 8/2010.


This dissertation explores the automatic generation of gestural score based control structures for a three-dimensional articulatory speech synthesizer. The gestural scores are optimized in an articulatory resynthesis paradigm using a dynamic programming algorithm and a cost function which measures the deviation from a gold standard in the form of natural speech production data. This data had been recorded using electromagnetic articulography, from the same speaker to which the synthesizer's vocal tract model had previously been adapted. Future work to create an English voice for the synthesizer and integrate it into a text-to-speech platform is outlined.

Weitere Links

Diss_Steiner_korr.pdf (pdf, 18 MB )

Deutsches Forschungszentrum für Künstliche Intelligenz
German Research Center for Artificial Intelligence