| |
Abstract:
In the speech production system, the lung and trachea are
connected to the glottis. The lungs are regarded as many sound
tubes with branches, thus the lung is supposed to have complex
characteristics. But the lung and trachea were ignored in many
kinds of speech synthesizers of vocal-tract analog. As improved
methods, several synthesizers were proposed to approximate the
characteristics of the lung and trachea. Since the effect of lung
and trachea was not correctly observed in the speech synthesized
by those synthesizers, it was difficult to evaluate the effect of
lung and trachea efficiently using the speech synthesizer of
vocal-tract analog. We propose a method to construct a speech
synthesizer with well approximation of the characteristics of the
lung and trachea. In our method, an equivalent small sound tube
is introduced as to separate the lung and the glottis, and the
lung and the glottis are modeled independently. Thus the proposed
method is able to be constructed easily using digital filters,
and the component of glottis or lung will be replaced easily. By
using the proposed method, it is possible to evaluate influences
of the lung impedance in the synthesized speech signal. In this
paper, detail of the proposed method is shown. To evaluate the
proposed method, several experiments of synthesis of speech
signal are performed by using the conventional and proposed
methods, we discuss about those speech signals, and show that our
method is useful to evaluate the effect of the lung and
trachea.
|
