Sixth International Conference on Spoken Language Processing (ICSLP 2000)

Beijing, China
October 16-20, 2000

Are Static MRI Measurements Representative of Dynamic Speech? Results From a Comparative Study Using MRI, EPG and EMA

Olov Engwall

Centre for Speech Technology (CTT), KTH, Stockholm, Sweden

As the acquisition times of static MRI have diminished, the focus of MRI studies has shifted from isolated phonemes to e.g. vowel-consonant sequences. The articulations are hence somewhat closer to running speech and measurements of contextual influence can be made. The acquisition in the vast majority of the MRI studies still requires artificially sustained articulations, however, and it is hence not evident that the measurements are representative of the subjectís normal speech. In order to assess the influence of the artificial sustaining, results from two coarticulatory studies, using static MRI and EMA-EPG, respectively, are compared.

The differences between the two studies are substantial concerning jaw position, lip protrusion and tongue contours. However, rather than showing non-representative articulations, the articulations in the MRI data are judged to represent a case of hyperarticulated speech, and should still be valid if considered as that.

Full Paper

Bibliographic reference.  Engwall, Olov (2000): "Are static MRI measurements representative of dynamic speech? results from a comparative study using MRI, EPG and EMA", In ICSLP-2000, vol.1, 17-20.