Department of Audiology and Speech Pathology, University of Tennessee Health Science Center, Knoxville, Tennessee.
J Voice. 2020 Mar;34(2):302.e1-302.e13. doi: 10.1016/j.jvoice.2018.09.012. Epub 2018 Oct 11.
OBJECTIVE/HYPOTHESIS: Using stimuli produced by classically trained female singers, this study sought to examine how inexperienced listeners' ratings of timbre dissimilarity relate to singer categorization, both between and within voice category.
METHOD/DESIGN: Stimuli were produced by six mezzo-sopranos and six sopranos classically trained female voice students at three pitches, C4, G4, and F5. Within each category, mezzo-soprano and soprano, three singers were classified as lighter in timbre and three were classified as heavier in timbre. At each pitch, the 12 sung stimuli were combined in all possible pairs and presented to 56 inexperienced listeners who judged each pair for the degree of timbre dissimilarity using a visual analog scale. All stimuli also were rated by expert voice professionals in two dimensions: (1) voice category and (2) vocal weight.
Multidimensional scaling (MDS) dimensions agreed with teacher-determined voice category at C4 and teacher-determined vocal weight at G4 and F5. MDS dimensions also correlated with single-note ratings of voice category and vocal weight provided by expert listeners who were unfamiliar with the singing voices used in this study. With limited exceptions, no single acoustic variable predicted the MDS dimensions that resulted from the inexperienced listeners ratings of timbre dissimilarity; however, weighted linear combinations of acoustic variables were highly predictive of numerous MDS dimensions.
In spite of a lack of both range and tessitura information, timbre dissimilarity ratings provided by inexperienced listeners yielded MDS dimensions that accurately predicted voice categories and vocal weight and that highly correlated to expert listeners ratings of these two parameters.
目的/假设：本研究使用经过经典训练的女性歌手所产生的刺激物，旨在探讨未经训练的听众对音色差异的评分如何与歌手分类相关，包括在和不在声部内。
方法/设计：刺激物由六名女中音和六名女高音古典声乐学生在三个音高 C4、G4 和 F5 处产生。在每个声部中，女中音和女高音中，有三位歌手被归类为音色较轻，有三位被归类为音色较重。在每个音高处，将 12 个演唱刺激物以所有可能的组合组合在一起，并呈现给 56 位未经训练的听众，他们使用视觉模拟量表对每对音色差异的程度进行判断。所有刺激物也由两位专家声乐专业人员在两个维度上进行评分：（1）声部和（2）声音重量。
多维尺度（MDS）维度在 C4 处与教师确定的声部一致，在 G4 和 F5 处与教师确定的声音重量一致。MDS 维度还与专家听众对声音类别和声音重量的单音符评分相关，这些专家听众不熟悉本研究中使用的演唱声音。除了一些有限的例外情况外，没有单一的声学变量可以预测未经训练的听众对音色差异的评分所产生的 MDS 维度；然而，声学变量的加权线性组合高度预测了许多 MDS 维度。
尽管缺乏音域和音域信息，未经训练的听众对音色差异的评分产生的 MDS 维度准确地预测了声部和声音重量，并且与专家听众对这两个参数的评分高度相关。
