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.
54 lines
1.4 KiB
54 lines
1.4 KiB
The Klatt formant synthesizer
|
|
|
|
purpose use as a strictly parallel synthesizer (Figure 12-4b). The experimenter
|
|
must decide beforehand which configuration is to be employed. The change in
|
|
configuration depends on the state of a single switch, and the program is smart
|
|
enough to avoid performing unnecessary computations for resonators that are not
|
|
used. To the extent possible, the synthesizer has been adjusted so as to generate
|
|
about the same output waveform whether the cascade/parallel configuration or the
|
|
all-parallel configuration is selected.
|
|
|
|
VOICING
|
|
SOURCE LARYNGEAL
|
|
D TRANSFER FUNCTION
|
|
|
|
(CASCADE)
|
|
ASPIRATION RADIATION
|
|
SOURCE D CHARACTERISTIC
|
|
FRICATION
|
|
|
|
TRANSFER FUNCTION
|
|
FRICATION
|
|
|
|
SOURCE
|
|
|
|
(PARALLEL)
|
|
|
|
OUTPUT
|
|
SPEECH
|
|
|
|
VOICING
|
|
SOURCE
|
|
|
|
I
|
|
RISTI
|
|
SOURCE (PARALLEL) CHARACTERISTIC
|
|
|
|
ouTPUT
|
|
SPEECH
|
|
|
|
FRICATION
|
|
SOURCE
|
|
Figure 12-4: Cascade/parallel configurations supported by MITalk
|
|
|
|
12.1.4 Waveform sampling rate
|
|
|
|
Most of the sound energy of speech is contained in frequencies between about 80
|
|
and 8000 Hz (Dunn and White, 1940). However, intelligibility tests of band-pass
|
|
filtered speech indicate that intelligibility is not measurably changed if the energy
|
|
in frequencies above about 5000 Hz is removed (French and Steinberg, 1947).
|
|
Speech low-pass filtered in this way sounds perfectly natural. Thus we have
|
|
selected 10,000 samples per second as the digital sampling rate of the synthesizer.
|
|
|
|
127
|