| |
Abstract:
The purpose of this paper is to introduce an articulatory
model based on cascaded first-order systems and to investigate
the articulatory dynamics and timing. Articulation associates
with complex process of muscle activation originating from motor
commands which are planned in the central nervous system.
Although its process has a complex transfer function, there is
advantage in simplifying it if possible. Although consisting of
simple components, the cascade model can be used for describing
approximately the transfer function of a complex system. In this
study, the flexibility of the cascaded first-order systems was
utilized for the description of the trajectory of articulatory
movements. In order to model articulator movements, the time
pattern of the impulse response of the model was fitted to the
velocity pattern of observed movements. Parameters describing the
model(time constant T, amplitude A, and input instant Tin of a
hypothetical motor command to the model) were obtained by
minimizing differences in velocity patterns between the model and
observed data. A combination of the responses originating from
the sequence of the hypothetical motor commands was used for
modeling successive articulatory movements. Tongue and jaw
movement data were collected with the magnetic and optical
sensing systems developed by the authors when a Japanese male
subject produced the utterances /eCeCe/, C=/p, t, k/ at normal
and fast speaking rates. Quantitative analysis based on the model
was performed by evaluating its parameters. The time interval
between the time instant Tin and the occurrence of related speech
phenomena like explosion of consonant C and onset of vowel /e/
was evaluated. In the case of /-etete-/, the experimental results
at the normal speaking rate showed that the jaw started to move
upward earlier than the tongue tip for the production of stop
consonant /t/. On the other hand, both articulators started to
move downward at the same time for the production of vowel /e/.
Evaluating model parameters amplitude A and time constant T
provided us with useful information on the differences in the
effects of speaking rate between primary articulator and
secondary articulator jaw. In the cases of /-etete-/ and
/-epepe-/, the primary articulators showed extreme reduction in
amplitude A. On the other hand, in the case of /-ekeke-/, the jaw
showed extreme reduction in parameter A. Since peak velocity for
individual ascending or descending movement is theoretically
proportional to the ratio of A to T , behavior of peak velocity
was also evaluated.
|
