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.
46 lines
1.2 KiB
46 lines
1.2 KiB
The phonetic component
|
|
|
|
Second formant transition for ow
|
|
|
|
1100
|
|
850
|
|
|
|
Frequency (kHz)
|
|
|
|
o R e
|
|
[0 T
|
|
|
|
0 200 230 380 460 800
|
|
Time (ms)
|
|
Second formant transition for GG ow
|
|
2 =
|
|
|
|
1448=1700+0.42(F2 -1700)
|
|
|
|
850
|
|
|
|
Frequency (kHz)
|
|
|
|
o 3 ]
|
|
0 200 280 380 460 800
|
|
|
|
Time (ms)
|
|
|
|
Figure 11-5: Synthesis strategy for a CV syllable
|
|
|
|
ways to improve the consonants and plan to extend this work to clusters and to
|
|
postvocalic allophones.
|
|
|
|
How can the intelligibility of the consonant-vowel syllables be further im-
|
|
proved? Of particular interest are two questions: 1) will the generalizations con-
|
|
cerning vowel categories and the form of the locus equations have to be modified,
|
|
or is it only necessary to modify the numbers that go into the equations, and 2) is a
|
|
spectral-matching procedure' of the type outlined above sufficient for the purpose
|
|
of intelligibility optimization? The generalizations that have been made will prob-
|
|
ably not have to be modified or abandoned. Each time that the perceptual data
|
|
have indicated a problem with a particular CV syllable or class of syllables,
|
|
reinspection of the spectral match between the synthesis and the average spectra
|
|
for talker DHK has revealed fairly substantial differences, and correction of the
|
|
|
|
115
|