The concept of a typical vowel sound is based on the variability of vowel pronunciation among men and women with different regional pronunciation and other variable characteristics. A common way of exploiting the information embodied in formant frequency is to represent each vowel by a centroid in the formant space with the realization that the centroid at best represents average behavior and does not represent variability across speakers. We set up an coordinate system using the formant frequencies and of word utterances. It is shown in Figure 4.2. Formants are extracted through an LPC-based peak-picking estimation method.
VOWEL | SAMPLES | ||
/a/ | 917 | 753 | 1385 |
/i/ | 805 | 315 | 2172 |
/u/ | 910 | 367 | 1332 |
/e/ | 765 | 500 | 1947 |
/o/ | 909 | 520 | 997 |
図 4.2: loci of Japanese five basic vowels measured
from the word utterances by native male speakers
A table of average formant frequencies is given in Table 4.3. Accordingly, we can set up the articulatory vowel diagram, as shown in Figure 4.6. It has been effective for a learner to correct his incorrect vowel pronunciation along with referring the tongue position of his vowel sound relative to the typical loci of the target vowel.