MULTIDIMENSIONAL ANALYSIS OFABNORMAL VOCAL QUALITY

对评估发声障碍的定量方法的需求已经被激发 对病理性发声的声学相关性进行了大量研究 质量。大多数基于声学的研究的目标是找到声学 测量：(1)可用于喉部病理检测 和(2)与感知程度和/或感知类型有很好的相关性 发音困难。例如，研究表明，语音样本来自 发音困难的扬声器表现出异常大的周期到周期的变化 基频(抖动)和语音幅度(微光)。它还 已经证明，发音困难的声音显示出异常大量的 加性噪声；即信噪比比 由正常扬声器进行的测量。然而，相对较少的是 已知单个声学参数如何组合以影响 听者对发声质量的总体印象。目前的研究是 旨在研究四个人的个人和互动贡献 控制感觉沙哑的声学参数：(1)抖动，(2) 抖动，(3)信噪比和(4)平均基频 (F)。实验室的计算机将被用来合成元音 抖动、抖动、信噪比(S/N)和平均值的变化 基频(FO)。这些刺激物将用于一系列 心理声学实验主要是为了回答两个问题：(1) 被感知的个体或单变量之间的关系是什么 声音沙哑和抖动，微光和S/N比，(2)如何抖动， 微光、S/N比值和平均FO共同影响感知声音嘶哑。这个 多变量听力实验的结果将使用 称为多维尺度(MDS)的统计过程。MDS分析 将被用来确定：(1)潜在的主观维度 感知到的沙哑，(2)给予每个人的相对知觉权重 维度和(3)这些维度与四个维度之间的关系 声学特性被操纵。然后，将尝试 将这些发现扩展到声音嘶哑评级和声学测量 从自然产生的正常和无序的声音中获得的。这个 这些实验的结果应该被证明对开发一种 基于听觉的指数，对数量和质量都很敏感 声音沙哑。这种类型的测量技术将是有用的 为临床医生和研究这一关系的科学家配音 在感知的语音质量和语音的声学属性之间。

The need for quantitative methods to evaluate voice disorders has motivated a good deal of research into the acoustic correlates of pathological vocal quality. The goal of most acoustic-based research is to find acoustic measurements that: (1) can be used in the detection of laryngeal pathology and (2) correlate well with perceived degree and/or perceived type of dysphonia. For example, it has been shown that voice samples from dysphonic speakers exhibit abnormally large cycle-to-cycle variations in fundamental frequency (jitter) and voice amplitude (shimmer). It has also been demonstrated that dysphonic voices show unusually large amounts of additive noise; i.e. signal-to-noise ratios are poor compared to measurements made from normal speakers. However, relatively little is known about how individual acoustic parameters combine to influence a listener's overall impression of vocal quality. The present research was designed to study both the individual and interactive contributions of four acoustic parameters in controlling perceived hoarseness: (1) jitter, (2) shimmer, (3) signal-to-noise ratio and (4) mean fundamental frequency (FO). A laboratory computer will be used to synthesize sustained vowels varying in jitter, shimmer, signal-to-noise (S/N) ratio and mean fundamental frequency (FO). These stimuli will be used in a series of psychoacoustic experiments designed primarily to answer two questions: (1) what are the individual or univariate relationships between perceived hoarseness and jitter, shimmer and S/N ratio, and (2) how do jitter, shimmer, S/N ratio and mean FO combine to affect perceived hoarseness. The results of the multivariate listening experiments will be analyzed using a statistical procedure called multidimensional scaling (MDS). MDS analysis will be used to determine: (1) the subjective dimensions underlying perceived hoarseness, (2) the relative perceptual weight given to each dimension and (3) the relationship between these dimensions and the four acoustic properties being manipulated. An attempt will then be made to extend these findings to hoarseness ratings and acoustic measurements obtained from naturally occurring normal and disordered voices. The results of these experiments should prove very useful in developing an acoustically based index that is sensitive to both the quantity and quality of hoarseness. A measurement technique of this type would be useful both to voice clinicians and to voice scientists studying the relationship between perceived voice quality and the acoustic attributes of speech.