4th International Conference on Spoken Language Processing
Philadelphia, PA, USA
This study applied the uniform scaling method (Nordstroem and Lindblom, 1975) within and across the two languages to the formant data collected equivalently from 40 healthy subjects, which formed four groups of 10 subjects each: Korean males, Korean females, American males, and American females. Then, the formant values were converted to a perceptual unit, mel, and plotted on the first formant against the second formant axes. From the cross-language comparison we observed that the vowels are placed to maintain sufficient perceptual contrast within each vowel system which supports the notion of Lindblom's (1990) "sufficient perceptual contrast". There were greater cross-language differences for the vowels [u] and [a] than for the others. If Korean [u] has a relatively low F2 value it might be confused with Korean [I]. Therefore, Korean [u] has a relatively low F2 value to keep sufficient perceptual distance from [i] and [u]. The AE speakers separate sufficiently the two low vowels [a] and [oe] by around 400 mel. On the other hand, Korean [a] is placed at the corner of a regular triangle formed by acoustically neighboring vowels [E] and [U] simply to maintain sufficient contrast because there're no competitive vowels along the F2 dimension. Similar perceptual contrast is maintained between the AE tense and lax vowels. Lax vowels are pushed inside the AE vowel space to secure sufficient perceptual distance to adjacent vowels. This way each vowel system seems to maintain sufficient perceptual contrast.
Bibliographic reference. Yang, Byunggon (1996): "Perceptual contrast in the Korean and English vowel system normalized", In ICSLP-1996, 114-117.