Nábĕlek A K, Czyzewski Z, Crowley H
Department of Audiology & Speech Pathology, University of Tennessee, Knoxville 37996-0740.
J Acoust Soc Am. 1994 May;95(5 Pt 1):2681-93. doi: 10.1121/1.409837.
Location of boundaries (the 50% response point) and slopes of identification functions were determined for synthesized /a-aI/ vowel continua. Within each continuum, the stimuli contained a steady-state segment followed by a transition in which the frequencies of formants changed in time. Here, F1 changed in a downward direction and F2 changed in an upward direction. Total duration of each stimulus was 200 ms. The duration of the transition was increased in steps from 0 to 140 ms. Two patterns of formant transition were used: (1) formants changing in the direction of, but not reaching, target frequencies (except in the end-point stimulus), and (2) formants reaching F1 and F2 targets. The data were collected with ten normal-hearing and ten hearing-impaired subjects. The boundaries and slopes were determined for four listening conditions: quiet, noise, short reverberation (0.8 s), and long reverberation (1.1 s). The location of boundaries depended upon: (1) pattern of formant transitions, (2) listening condition, and (3) status of subjects' hearing. Generally, longer transitions were needed for formants changing in the direction of, but not reaching, target frequencies, than for those reaching F1 and F2 targets. The required transition durations were similar in quiet and noise, but were longer in reverberation. The hearing-impaired subjects generally required longer transitions to reach the boundaries than normal-hearing subjects. The slopes of the identification functions were shallower in either noise or reverberation than in quiet and were shallower for hearing-impaired than for normal-hearing subjects. In reverberation, the slopes for formants reaching targets were shallower than the slopes for stimuli with formants changing in the direction of target frequencies. The relationships between these findings and identification errors for naturally produced tokens of the diphthong /aI/ are discussed.
针对合成的/a - aɪ/元音连续体，确定了边界位置（50%反应点）和识别函数的斜率。在每个连续体中，刺激包含一个稳态段，随后是一个过渡段，其中共振峰频率随时间变化。在此，F1向下变化，F2向上变化。每个刺激的总时长为200毫秒。过渡段的时长从0到140毫秒逐步增加。使用了两种共振峰过渡模式：（1）共振峰朝着目标频率方向变化但未达到目标频率（端点刺激除外），以及（2）共振峰达到F1和F2目标频率。数据收集自十名听力正常和十名听力受损的受试者。针对四种聆听条件确定了边界和斜率：安静、噪声、短混响（0.8秒）和长混响（1.1秒）。边界位置取决于：（1）共振峰过渡模式，（2）聆听条件，以及（3）受试者的听力状况。一般来说，共振峰朝着目标频率方向变化但未达到目标频率时，比达到F1和F2目标频率时需要更长的过渡时间。在安静和噪声环境中所需的过渡时长相似，但在混响环境中更长。听力受损的受试者通常比听力正常的受试者需要更长的过渡时间来达到边界。在噪声或混响环境中，识别函数的斜率比在安静环境中更平缓，且听力受损受试者的斜率比听力正常受试者更平缓。在混响环境中，共振峰达到目标频率时的斜率比共振峰朝着目标频率方向变化的刺激的斜率更平缓。讨论了这些发现与双元音/aɪ/自然发音样本的识别错误之间的关系。
