Towards a speakerindependent representation of tongue-posturing for speech”, in Speech motor control in normal and disordered speech, : B. Maassen et al., Eds

Abstract. In previous work the three-mode factor analysis technique PARAFAC had given a revealing picture of tongue configurations for vowels derived from both EMMA and mid-sagittal NMRI data. While very parsimonious and elegant, it appeared, however, that the PARAFAC model might be too restrictive to capture speaker-specific effects in consonantal (from EMMA) and non-midline (from NMRI) aspects of articulation. In the current work we are exploring more general statistical models. To make NMRI data amenable to the statistical techniques we outline a jaw-based registration method followed by extraction of intrinsic tongue-surface coordinates on a spherical grid. 1