11. A generative synthesis would recreate this word moment by moment.
Using the formants
drawn on this spectrogram as a guide to which tones are the most prominent in the natural
production of “hide,” which 5 tones would a generative synthesis program use to recreate
this word?
Notice that we’re using a sampling rate of 100 ms.
Lowest Tone
Tone 2
Tone 3
Tone 4
Highest Tone
300 ms
900
1100
2100
3400
4800
400 ms
950
1050
2100
3400
4800
500 ms
600 ms
700 ms
800 ms
900 ms
12. How many data points do you have for the phoneme /h/?
For /ɑ/?
Is Generative Synthesis
sensitive to phoneme boundaries?
13. The word above is about 1 second long.
It is not uncommon that a Generative Synthesis
program would take 16,000 samples each second.
Given that the human production has 40
harmonics at any given moment, how many tones would be used to synthesize this word?
14. There are three ways to improve the synthesis you outlined in (17).
Can you name all three?
a.
b.
c.
15. What if you wanted to synthesis the word “dye” [dɑi] instead using generative synthesis.
What would you have to do?
16. What would you have to do if you wanted to synthesize the word “hide” in a Chicago dialect
instead of a Southern California dialect?
