Modeling Subject-Specific Formant Transition Patterns in / / Sequences

Careful observation reveals that the formant transition patterns in a particular VCV sequence vary according to subjects. It is the case in /a a/ uttered by French speakers. This variation can be explained by the realization of acoustically equivalent but differently articulated / / or /a/ segments and their combination. This paper reports an acoustic simulation experiment aiming at replicating this phenomenon. The results are compared to simultaneous ultrasound tongue contour data and discussed in the perspective of speech inversion.