多次元尺度構成法による無喉頭者音声の声質の分析

概要

論文の詳細を見る
This study investigates and acoustically defines some of the perceptual parameters which characterize alaryngeal voice quality. Twenty voice samples of vowel /e/ were recorded from 16 alaryngeal and four normal speakers, and were paired, randomized and presented to 20 listeners with normal hearing. The listeners rated the dissimilarities of all combinations of the voices, and these dissimilarities were analyzed by a nonmetric multidimensional scaling method called SMACOF. Five acoustical measures and 21 psychoacoustical ratings were made for each vowel, and these provided the interpretation for the SMACOF solution. The results indicate that at least two dimensions are necessary to describe the alaryngeal voice quality. One is connected with the listeners' preferences for the voices and differentiates the alaryngeal voices from the normal ones. This dimension correlates with the fundamental frequency and the relative spectral level of the high frequency components. Another dimension indicates the abnormalities peculiar to the voices produced with artificial larynges and differentiates them from the normal and esophageal voices. It correlates positively with periodicity in the excitation signal waveform and negatively with aperiodicity in the pitch or in the amplitude. These results suggest some direction toward the enhancement of the alaryngeal voice quality.
社団法人日本音響学会の論文