Chen M Y
Department of Electrical Engineering and Computer Science, Massachusetts Institute of Technology, Cambridge 02139, USA.
J Acoust Soc Am. 1995 Nov;98(5 Pt 1):2443-53. doi: 10.1121/1.414399.
One of the more prevalent abnormalities that contribute to reduced intelligibility in the speech of hearing-impaired speakers is inadvertent nasalization. Acoustic analysis of the speech of hearing-impaired children and nasalized vowels of normal-hearing adults indicate the presence of an extra pole-zero pair between the first and second formants and a reduced first-formant prominence. The difference between the amplitude of the first formant and the amplitude of the extra peak, A1-P1, was shown to be a promising measure that correlates with listener judgments of the degree of vowel nasality in utterances of hearing-impaired and normal-hearing children. To obtain further validation of these parameters as measures of nasality, A1 and P1 were systematically manipulated in synthetic utterances. The results of perceptual experiments with a number of synthesized words (of the form bVt) showed somewhat different relative importance of the two parameters, depending on the vowel. High correlation of A1-P1 with the average nasality perception judgments was found for each of the stimuli except those for which the first formant was very close to the extra peak frequency. A procedure was developed for correcting the value of A1-P1 to normalize for different vowels, based on the frequencies of the first two formants.
