| |
Abstract:
EMA and EPG offer complementary insights into tongue
articulation. In this experiment simultaneous EMA and EPG data
were recorded for high vowel continua produced by one speaker.
The data were examined in order to estimate the location of the
lingual constriction. The aim of the experiment was to correlate
such articulatory estimates of constriction location with formant
measurements in a way that would allow comparison with
model-based nomograms in Fant [1]. The focus in this paper
however is on the exploitation of the two techniques. It was
envisaged that their simultaneous use might enhance the analysis
possible from either alone and that one might be used as a check
on the accuracy of the other. A tongue circle model was used to
estimate the constriction location from the EMA. The narrowest
point of the constriction was deemed to be the point where the
circle was closest to the roof of the mouth. These measures
correlated well with estimates of the lingual constriction mad! e
from scaled vertical views of the EPG palate for pre- to
post-palatal articulations. Approximations of the midsagittal
outline of the tongue were made on the basis of the two types of
data and compared. In the case of EPG these were drawn on the
basis of the innermost activated electrode on each EPG row, when
the palate is viewed from the side. For the EMA data the outline
was based on the circle model. Results suggest that the EPG-based
outline frequently underestimates tongue height, particularly for
more retracted articulations. The main reason appears to be that
EPG data involve lateral tongue contact whereas EMA data relate
to the midsagittal line.
|
