| |
Abstract:
Although existing electromagnetic mid-sagittal articulography
Systems have provided invaluable information on speech
kinematics, they are not suited to monitor substantial lateral
deviations of the tongue from the midline without error. Due to
their need for a fixed head-position they also restrict the
subjects' freedom of movement and thus potentially compromise the
naturalness of their speech. Accordingly we started the
development of a new electromagnetic articulography system at the
IPSK, which is able to acquire data within a spherical area
without further restrictions. The new 3-D-EMA system is based on
a spherical placement of 6 transmitter coils but operates with
the same receiver coils as the Carstens AG100 system. We
significantly reduced the number of analog components and built a
digital-featured device. During the last couple of months we did
several 3-D-Measurements to estimate the measurement accuracy of
the new device. We obtained a spatial resolution better than 1 mm
and a rotation detection with approx. 1 degree accuracy. We were
also able to verify that the magnetic field strength of the new
device is about 2 μT at the center of the measurement area
which is comparable to former model calculations of magnetic
field strength. We present first 3-D measurement data on a human
subject collected with a prototype of the new 3-D-EMA system. To
compensate movements of the subjects head during the measurement
and to transform the measured data into a skull-fixed coordinate
system, reference coils must be placed on the subjects head and
an extra algorithm has to subtract head-movements from the
movements of the sensors. Although we had to deal with some
technical problems caused by preliminary parts of our prototype,
the data collected to date shows that the new 3-D system is
capable to aquire multi-dimensional data of speech movements.
Especially the limited measurement frequency of the prototype (25
Hz) prevented us from collecting more articulographic significant
data, since with only a few measured points per utterance it is
difficult to analyze trajectories of speech movement. Still, even
this prototype shows promising prospects and a new set of
hardware components is currently being assembled. In the near
future we will focus on the calibration procedure and - with
increased number of channels - on the compensation of head
movements during the measurements with sufficient accuracy to
obtain more phonetically relevant data.
|
