Emanuel F, Scarinzi A
J Commun Disord. 1979 May;12(3):263-72. doi: 10.1016/0021-9924(79)90047-9.
The effects of vocal register (vocal fry, modal, and falsetto) on the relative roughness and spectral noise level (SNL) of isolated test vowels (/u/ and/ae/) were investigated. Each of the three vocal registers; the obtained samples were magnetically recorded. A panel of 11 listeners independently verified by perceptual judgments that each recorded sample was produced in the "desired" vocal register. Also, they independently rated (on a 5-point equal-appearing intervals scale) the roughness of each test production. A median of the 11 roughness ratings (MRR) available was then obtained as an index of the roughness of each vowel sample. Additionally, a narrow-band (3 Hz) acoustic spectrum of each test production was made in which measures of vowel spectral noise were obtained. A mean over 25 noise measures per test production (from 100 to 2600 Hz) served as an index of the spectral noise level (SNL) for each vowel sample. The major finding in the study was that the MRR and the SNL for productions of both test vowels diminished significantly across vocal registers, from fry, to modal, to falsetto.
研究了声区(喉塞音、正常发声和假声)对孤立测试元音(/u/和/ae/)的相对粗糙度和频谱噪声水平(SNL)的影响。三个声区中的每一个;所获得的样本都进行了磁记录。由11名听众组成的小组通过感知判断独立验证每个录制样本是在“期望的”声区产生的。此外,他们独立地(在5点等距量表上)对每个测试发音的粗糙度进行评分。然后获得11个粗糙度评分(MRR)的中位数,作为每个元音样本粗糙度的指标。此外,对每个测试发音制作了一个窄带(3Hz)声谱,从中获得元音频谱噪声的测量值。每个测试发音的25个噪声测量值(从100到2600Hz)的平均值作为每个元音样本频谱噪声水平(SNL)的指标。该研究的主要发现是,两个测试元音发音的MRR和SNL在声区之间显著降低,从喉塞音到正常发声,再到假声。