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.
This file contains ambiguous Unicode characters that may be confused with others in your current locale. If your use case is intentional and legitimate, you can safely ignore this warning. Use the Escape button to highlight these characters.
From text to speech: The MITalk system
N
I
X 6 6
3
& 4 > 4
o
=
o
o 2 2
©
= 2 4 6 8 2 4 6 8 2 4 6 8
F1 target of vowel (kHz) F1 target of vowel (kHz) F1 target of vowel (kHz)
= 20 2.0 2.0
g "
E 1.6 / 1.6 / 1.6 *
g Xr{ /
2 1.2 1.2 1.2
=
3]
© /
>
= 8 8 8
N
w
8 12 16 20 8 12 16 20 8 12 16 20
F2 target of vowel (kHz) F2 target of vowel (kHz) F2 target of vowel (kHz)
= 3.0 3.0
I
<
% 25 2.5 =
8 ‘ ‘
o
220 2.0
L
O
>
= 15 1.5
™
.
15 20 25 8.0 15 20 25 3.0 15 20 25 3.0
F3 target of vowel (kHz) F3 target of vowel (kHz) F3 target of vowel (kHz)
Figure 11-4: Frequency of the lowest three formants measured at voicing onset
for syllables involving BB, DD, and GG
ticular software synthesizer (Klatt, 1980; see Chapter 12), but perhaps future publi-
cation of the numbers would be of some value to those who wish to implement the
synthesizer program.
11.2.3 Intelligibility evaluation
The intelligibility of CV syllables produced by the rules was evaluated by syn-
thesizing 336 different CVt syllables in a random order. The tape was played to
five phonetically trained listeners who transcribed both the consonants and the
vowels. The vowel identification rate was 99 percent and the consonant identifica-
tion rate was 95 percent. While these results are encouraging, we continue to seek
114
