You can not select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
44 lines
1.5 KiB
44 lines
1.5 KiB
From text to speech: The MITalk system
|
|
|
|
800 Back «<—> Front
|
|
700
|
|
)
|
|
< 600
|
|
-
|
|
(o
|
|
e
|
|
5
|
|
e 500
|
|
S
|
|
2
|
|
ic
|
|
400
|
|
300
|
|
|
|
800 1000 1200 1400 1600 1800 2000 2200
|
|
Second formant frequency
|
|
|
|
Figure 11-2: First and second formant motions in English vowels
|
|
|
|
Figure 11-4 contains a plot of average frequencies for the lowest three for-
|
|
mants, as measured at voicing onset following plosive release in syllables contain-
|
|
ing BB, DD, and GG. Values before 16 vowel nuclei are plotted on the vertical axis
|
|
as a function of the formant frequency seen in the early part of the vowel. Notice
|
|
that, for the first formant, data can be well approximated by a straight line.
|
|
However, the value of F1 at voicing onset, i.e., when F1 is first perceptible, is as
|
|
high as 500 Hz before low vowels because most of the rapid rise in F1 at release
|
|
takes place prior to voicing onset.
|
|
|
|
If formant values at voicing onset fall on a straight line in this kind of plot, a
|
|
“locus theory” description of the data is possible; that is, there exists a locus theory
|
|
equation with two free parameters that can predict F1 at voicing onset from a
|
|
knowledge of the vowel target. An example of this synthesis procedure will be
|
|
presented shortly.
|
|
|
|
Before describing aspects of the synthesis strategy, it should be noted that
|
|
there exist articulatory motivations for dividing vowels into these three sets. An-
|
|
ticipatory coarticulation of the vowel features “front-back” and “rounded-
|
|
unrounded” can explain all of the acoustic observations noted in the figures.
|
|
|
|
112
|