Computational Biology and Language - Ambient Intelligence for Scientific Discovery - page 35

Fig. 7. Spectrograms of same utterances between different speakers and different ut-
terances by same speaker: X-axis shows progression of time and y-axis shows different
frequency bands. The energy of the signal in different bands is shown as intensity in
grayscale values with progression of time. (A) and (B) show spectrograms of the same
sentence “How to recognize speech with this new display” spoken by two different
speakers, male and female. Although the frequency characterization is similar, the for-
mant frequencies are much more clearly defined in the speech of female speaker. (C)
shows the spectrogram of the utterance “How to wreck a nice beach with this nudist
play” spoken by same speaker as in (B). (A) and (B) are not identical even though they
are composed of the same words. (B) and (C) are similar to each other even though
they are not the same sentences. See text for discussion.

3.3 Speaker Variability

Consider the signal characteristics of a word spoken by two different persons,
especially if one is female and the other is male. Although the fundamental
nature of the sounds remains the same, the overall absolute values of signal
composition would be different. For example, a vowel sound would still have the
same periodic nature in both utterances, but the frequency would be different.
See for example, the frequency compositions of the same sentence spoken by a
male and female speaker shown in Fig. 7A and 7B.
The analogy of speaker variability in the protein world can be found in the
following broad categorization of proteins: the majority of a cell's proteins are