Marilyn Y. Chen
Res. Lab. of Electron. and Dept. of Elec. Eng. and Comput. Sci., MIT, Rm. 36-511, Cambridge, MA 02139
Acoustic analysis of nasalized vowels in the frequency domain indicates the presence of extra peaks, one between the first two formants with amplitude P1 and one below the first formant with amplitude P0; the first-formant amplitude A1 is also reduced relative to its amplitude for a nonnasal vowel. These acoustic characteristics may be explained by speech production theory. The objective of this study is to determine the base line values for the acoustic parameters in quantifying nasality. A1-P1 and A1-P0 were tested as measures of nasality by comparing vowels adjacent to nasal consonants and those adjacent to stops for English speakers. For high vowels, A1-P1 is a better parameter to distinguish nasal vowels from non-nasal vowels, with an average difference of 10 dB; however, A1-P0 is better for low vowels, with an average difference of 7 dB. French nasal vowels adjacent to stops were also examined. To further test these parameters as measures of nasality, they were systematically manipulated in synthetic speech and presented in perceptual experiments. High correlation was obtained between nasality perception and normalized acoustic parameters. [Work supported by NSF, NIH, and LeBel fellowship.]