Ryuta Kataoka
UNC Craniofacial Ctr., Univ. of North Carolina, Chapel Hill, NC 27514
Ken-ichi Michi
Showa Univ., Tokyo, Japan
Kaoru Okabe
Coretex Corp., Tokyo, Japan
To evaluate hypernasality quantitatively, frequency analyses were performed on the Japanese isolated vowel /i/ produced by sixteen cleft palate patients with or without speech appliances and four normal subjects. These 26 speech samples were rated by 20 listeners. Analysis of the one-third-octave power spectra revealed an increase in power level between the first and second formant and a reduction in second and third formant regions among hypernasal speech samples. Factor analysis of the five-point listener's ratings revealed that the consensus perception of hypernasality accounted for 75% of the total variance. Multiple regression analysis revealed a high correlation (0.82) between the perception of hypernasality and the peak levels of two clusters of 1/3-octave bands: the sixth, seventh, and eighth 1/3-octave multiples, and the tenth, eleventh, and twelfth multiples from the band containing the fundamental frequency. Based on these findings, a system for evaluating hypernasality was developed using a computer and 1/3-octave bandpass filter. This approach is presently being used to assess the relationship between velopharyngeal port area obtained from pressure-flow study and hypernasality evaluated by this system.