5th European Conference on Speech Communication and Technology

Rhodes, Greece
September 22-25, 1997

Acoustic Modelling of American English /R/

Carol Y. Espy-Wilson (1,5), Shrikanth Narayanan (2), Suzanne E. Boyce (1,3,5), Abeer Alwan (4)

(l) Dept. of Electrical and Computer Engineering, Boston University, Boston, MA, USA
(2) AT&T Labs, Murray Hill, NJ, USA
(3) Dept. of Communication Sciences and Disorders, University of Cincinnati, OH, USA
(4) Dept. of Electrical Engineering, University of California at Los Angeles, USA
(5)Research Laboratory of Electronics, Massachusetts Institute of Technology, Cambridge, MA, USA

The low F3 of American English /r/ (typical range 1300-1900 Hz) is accompanied articulatorily by constrictions in the pharyngeal, palatal and labial regions. Because acoustical theory predicts that formants will lower at points of maximum Volume velocity in the vocal tract, and because such points occur in the pharyngeal, palatal and labial regions, many investigators have speculated that the combination of these constrictions accounts for the low F3 of /r/. In this paper, we use the Maeda vocal tract modelling software to compare theoretical predictions of constriction location to data gathered from two American English speakers via Magnetic Resonance Imaging (MRI). We conclude that additional mechanisms are required to explain the acoustics of American English /r/.

Full Paper

Bibliographic reference.  Espy-Wilson, Carol Y. / Narayanan, Shrikanth / Boyce, Suzanne E. / Alwan, Abeer (1997): "Acoustic modelling of American English /r/", In EUROSPEECH-1997, 393-396.