Digital Signal Processing Reference
In-Depth Information
2,00
Frequency domain
1,50
1,00
0,50
0,00
0
250
500
750
1000
1250
1500
1750
2000
2250
2500
2750
3000
3250
3500
3750
Hz
Illustration 70:
Waterfall analysis: frequency-time landscape of the sequence „I owe you“
This is an Illustration which corresponds to Illustration 68 in the frequency domain. At the top there are
the spectra of the beginning of the word, at the bottom those of the end of the word. The "i" spectra at the
top and the „u“ spectra at the bottom are clearly identifiable. In the middle there are the "o" spectra with
a higher frequency than the "u" spectra and the blurred transition spectra.
An ideal frequency-time-landscape would differ slightly from the one presented here. It would be composed
of several partly parallel disc-shaped pinnacles stretching from the foreground into the background with
deep ravines between the peaks. This perfect frequency-time-landscape would present in an ideal way the
connection between length of time and bandwidth and thus the
UP
for this event. See also Illustration 73.
The way experienced lecturers and speakers stress a particular
word contributes to its intelligibility. They stress the vowels.
Those who know how to stress words well ought in theory to be
good singers as well because a stre-e-e-ssed word is basically a
sung word with a clear pitch.
As you will know, language consists of a sequence of vowels and consonants. Vowels are
near-periodic. But this is not true of consonants. They are like noises, some of which use
the noise of the air stream (try and pronounce a very long "s") some are like small explo-
sions, such as "b", "p", "k", "d", "t". These consonants have a continuous spectrum
without near-periodic elements.
The second example is the English word "history" because it consists clearer of an alter-
nating sequence of vowels and consonants.
