ISCA Archive PSP 2005
ISCA Archive PSP 2005

The potential role of speaker's vocal tract morphology onto speech perception: preliminary results

Susanne Fuchs, Jana Brunner, Pascal Perrier, Christian Geng, Bernd Pompino-Marschall

Perkell et al. (2003) have experimentally shown for American English native speakers relations between perceptual discrimination capacity and articulatory gestures accuracy: the best the discrimination, the more canonical and the less variable the articulation. This exemplifies how speech production and speech perception interact in their developments and how they contribute to shape each other. In the same vein, our work addresses the issue whether morphological characteristics of the vocal tract can influence perceptual patterns and discrimination. In a previous work on token-to-token variability during vowel production in German (Mooshammer et al. 2004), it was shown that for high/front vowels speaker dependent morphological differences of the palate shape influenced the articulatory variability. Indeed, for the subject having a flat coronal palate shape, tongue sensors located in the mid-sagittal plane exhibited very small, nearly circular dispersion ellipses, whereas two speakers with dome shaped palates showed distinctively larger variability patterns. In accordance with Perkell et al. (1997) we interpreted these results with respect to perception: for a flat palate small articulatory changes have a large impact onto the acoustics, in comparison with a dome shaped palate. Thus, in order to avoid perceptual confusion between neighboring vowels, speakers with a flat palate shape could control their tongue positioning more precisely. The current study aims at assessing the relation between vocal tract morphology and the structure of the perception space of the speakers. First, we investigate the relation between formants variability and palatal shape by means of simulations with a 2D biomechanical tongue model (see e.g. Perrier et al. 2003). Simulations were run for sequences /kak/, /kik/ and /kuk/ for a set of different muscle activations for each vowel (resulting in realistic variability of tongue positioning for each vowel) and for different palate shapes (the default palate of the model, a flat and a dome shaped palate). Area functions of the vocal tract (Perrier et al. JSLR, 1992). The five first formants were, then, computed, and finally the sound was synthesized with a Klatt Synthesizer. Results provide evidence in the predicted direction, i.e. changes in articulatory movements have a larger auditory impact when the flat palate shape, and it is especially strong for high/front vowels. In second phase we have run perceptual tests of these artificial stimuli with the three subjects analyzed in Mooshammer et al. (2004), and the link between perceptual discrimination and the shape of the palate was assessed.

Cite as: Fuchs, S., Brunner, J., Perrier, P., Geng, C., Pompino-Marschall, B. (2005) The potential role of speaker's vocal tract morphology onto speech perception: preliminary results. Proc. ISCA Workshop on Plasticity in Speech Perception (PSP 2005), 138-141

  author={Susanne Fuchs and Jana Brunner and Pascal Perrier and Christian Geng and Bernd Pompino-Marschall},
  title={{The potential role of speaker's vocal tract morphology onto speech perception: preliminary results}},
  booktitle={Proc. ISCA Workshop on Plasticity in Speech Perception (PSP 2005)},