McCaffrey H A, Sussman H M
University of Texas at Dallas.
J Speech Hear Res. 1994 Aug;37(4):938-51. doi: 10.1044/jshr.3704.938.
Vowel auditory formant distances were obtained from speakers with hearing loss to investigate how perceptual constraints affect the contrastiveness and intelligibility of their spoken vowels. These distances were evaluated in relation to the 3-Bark critical distance principle for vowel height and place as described by Syrdal (1985) and Syrdal and Gopal (1986). Seven speakers with profound hearing loss, 10 with severe hearing loss, and seven with normal hearing produced the vowels /u/, /i/, /I/, /ae/, /a/, and /--/ in an /hVt/ context. Vowel formants and fundamental frequencies were obtained with acoustic spectrographic and LPC analysis and converted to Bark values to establish auditory formant distances. Confusion matrices were constructed from normal listeners' identifications of recorded vowel productions. When frequency data were transformed to a Bark auditory scale, increasing convergence of vowel targets was obtained with increase in hearing loss. Percent correct identifications of the vowels produced by the three groups reflected speaker group differences seen in vowel contrastiveness/overlap in auditory phonetic space. Four levels of performance based on error incidence and type were determined. F1-F0 by F3-F2 Bark distance coordinate plots of a given speaker's vowel space reflected the differential intelligibility scores shown by confusion matrices of individual speakers from the four performance levels. Vowel organization by speakers with hearing loss was influenced by (a) formant critical distance, and (b) formant audibility. The least audible formants, F2 and F3, showed the greatest effects of severe and profound hearing loss. F1 and F0 showed further change with the most profound losses and revealed individual differences as well.
从听力损失患者那里获取元音听觉共振峰距离，以研究感知限制如何影响他们所发元音的对比度和可懂度。这些距离是根据Syrdal（1985年）和Syrdal与Gopal（1986年）所描述的关于元音高度和位置的3 Bark临界距离原则进行评估的。七名重度听力损失患者、十名重度听力损失患者和七名听力正常者在/hVt/语境中发了元音/u/、/i/、/I/、/ae/、/a/和/--/。通过声学频谱分析和线性预测编码（LPC）分析获得元音共振峰和基频，并将其转换为Bark值以确定听觉共振峰距离。根据正常听众对录制的元音发音的识别构建混淆矩阵。当频率数据转换为Bark听觉标度时，随着听力损失的增加，元音目标的收敛性增加。三组患者所发元音的正确识别百分比反映了在听觉语音空间中元音对比度/重叠方面的说话者组间差异。根据错误发生率和类型确定了四个表现水平。给定说话者元音空间的F1 - F0与F3 - F2 Bark距离坐标图反映了来自四个表现水平的个体说话者混淆矩阵所示的不同可懂度得分。听力损失患者的元音组织受（a）共振峰临界距离和（b）共振峰可听度的影响。最不易听见的共振峰F2和F3受重度和极重度听力损失的影响最大。F1和F0在极重度听力损失时显示出进一步变化，并且也揭示了个体差异。
