This paper takes an in-depth comparative study on the children with the disease causing hyponasality (allergic rhinitis, adenotonsillar hypertrophy) and normal children for perceptual, acoustic and instrumental objective evaluations to understand the acoustic characteristics based on its causes. This study also aims to investigate the severity classification as a surgery indicator.
In total, 35 children with adenotonsillar hypertrophy, 33 children with allergic rhinitis and 32 normal children from four years and four months(52 months) to nine years and eleven months(119months) were investigated for this research. Three vowels /a, i. u/, 0% of nasal consonant {bada} sentence, 50% of nasal consonant {dongmulwon}, 100% of nasal consonant {eumma} were used for this investigation. Tascam(dr05) recorder and nasometer were used for the voice recording.
One-way ANOVA analysis was resulted in followings.
First, there was a significant difference between two groups in shimmer-apq5 by Cepstrum variable and voice recording from {dongmulwon} and {eumma} sentence. {dongmulwon} sentence had significant differences by shimmer-apq11 which is a Cepstrum variable with nasometer voice recording, by shimmer_apq3 of vowel /a/ and by shimmer-ddp. Those two suggested that a group with allergic rhinitis and a group with adenotonsillar hypertrophy had greater shimmer values compared to a normal group (4.04/4.21).
Second, HNR(Harmonic to Noise Ratio) which is a vocal feature of nasometer recoding had a meaningful difference among three groups in vowel /a/. A normal group > a group with adenotonsillar hypertrophy > a group with allergic rhinitis had greater values in sequence.
Third, acoustic characteristics of voice variables from nasometer recording showed a meaningful difference from all those sentences. The sentences with 0% of nasal consonant {bada}, 50% of nasal consonant {dongmulwon} and 100% of nasal consonant {eumma} had significant differences between groups.
Fourth, the skewness of spectrum variable from Tascam recording of vowel /a/, the skewness and kurtosis from the spectrum starting point to the maximum amplitude of vowel /i/, the cepsSlope of {bada} and {eumma} from data in nasometer recording and cpp_m of vowel /a/ had the greatest value in a group with adenotonsillar hypertrophy(p<.05).
Fifth, auditory-perceptive evaluation rating scores of the data from Tascam recording and hyponasality from the date in nasometer recording corresponded each other. The group with adenotonsillar hypertrophy had the highest hyponasality.
To investigate the acoustic variables affecting groups and to evaluate the mirror-fogging test along with sound characteristics, the discriminant analysis was applied for the classification accuracy. The results are as follows.
First, {bada} of the data from Tascam recording and {eumma} from data in nasometer recording had 52% and the greatest discrimination variables were shim_apq5 and nNasalance.
Second, {eumma} and {bada} from recording data in nasometer had 51% of classification accuracy and the nNasalance was the greatest discrimination.
Third, nNasalance and mirror-fogging test results were analyzed to understand the influence on the nNasalance and mirror-fogging test to those three groups. It resulted in 30~41% of classification accuracy only with nNasalance, 60% of classification accuracy only with mirror-fogging test result. The classification accuracy with mirror-fogging test was 19~30% higher than with nNasalance.
Fourth, the nNasalance was the best variable which could well evaluate resonance disorders among all of the variables. Shimmer derived variables and moment variables such as skewness, kurtosis and standard deviation were proved to better classify the groups with resonance disorders in case Tascam(dr05) data.