Genetic Influence on Human Speech -Profiling One's Speech based on Perceptual and Acoustic Analyses

为了了解遗传和环境对人类语音的影响，对不同说话人群体中收集的标准语音样本进行了知觉和声学分析。进一步分析了选定的声学测量和知觉判断，以确定可以解释言语个体差异的关键特征1。对114名正常说话者(44名年轻人和70名老年人)的语音样本进行了分析，以考察年龄和性别对知觉和声学特征的影响。一些老年人说话的基频高于年轻人，而一些老年女性说话的基频低于年轻人。老年人的语速通常较慢，且有轻微的声音嘶哑。即使在老年人中，发音也是完好无损的(即没有发音错误和异常)。此外，一组成人说话者(N=70)的语音和声音样本被提供给成年听者用于性别和年龄估计。对正常说话者的性别和年龄进行了准确的估计，其中…结果1.有构音障碍的EAS患者(N=20)多为老年人。为了检验说话人内部的变异性，我们从年轻人(N=8)身上采集了10天(每天两次)的语音样本(口语朗读和长元音)。测量了语音和说话的基频、元音的共振峰频率、句子和单词的时长，发现与平均值的偏差约为10%，而不考虑说话者和测量方法。语音样本来自4-6岁幼儿(N=54)和23对双胞胎(N=46)。较年轻年龄组的词汇持续时间较长。对发声牙槽瓣/r/in//terebi/(日语电视)的声学分析表明，一些儿童出现典型的/r/共振峰运动，但在一些儿童的发音中发现了音隙(典型的发声牙槽停顿/d/、替换)和目标音素缺失(遗漏)。在20对双生子中，/r/in/terebi/的双生子发音符合率为90%(18/20)，分别基于正常发音(13对)和发音缺陷(5对)，不同发音对中的2对发音出现发音缺陷。根据从成人和儿童语音中准随机选择的语音样本(65个样本；58个说话者)的感知和声学特征来执行语音特征分析。从较高的共振峰频率估计的说话基频(即声调水平)、语速、声道长度是有效解释语音个体差异的关键特征。音调范围(可变性)、不流畅性和其他感知特征可用于描述和识别单个说话者。较少

To understand genetic and environmental influences on human speech, perceptual and acoustic analyses were performed on standard speech samples collected in various speaker groups. Selected acoustic measures and perceptual judgments were further analyzed to determine key features that would explain individual differences of speech.1. Speech samples of 114 normal speakers (44 young and 70 elderly adults) were analyzed to examine age and gender effects on perceptual and acoustic features. Speaking fundamental frequency was higher for some old man and lower for some old women than young adults. Speech rate was typically slower and mild hoarseness was noted in the elderly adults. Articulation was perceptually intact even in the elderly adults (i.e., free of misarticulation and abnormalities). In addition, a set of speech and voice samples of adult speakers (N=70) was presented to adult listeners for gender and age estimation. Gender and age were accurately estimated for normal speakers wher … More eas speakers with dysarthria (N=20) were mostly judged as older persons.2. To examine intra-speaker variability, speech samples (oral reading and prolonged vowels) were obtained from young adults (N=8) across 10 days (twice in each day). Voice and speaking fundamental frequency, formant frequencies of vowels, sentence and word duration are measured, and approximately 10% of variations from mean were found regardless of speakers and measures.3. Speech samples were collected from young children between age of 4 and 6 (N=54) and 23 twin pairs (N=46). Word duration was longer for younger age groups. An acoustic analysis of voiced alveolar flap /r/ in //terebi/ (television in Japanese) revealed that typical formant movement for /r/ were seen in some children, but stop gap (typically seen for voiced alveolar stop /d/, substitution)and missing target phone (omission) was found in productions of some children. Among 20 twin pairs, consistency of twin-pair articulation of /r/ in /terebi/ was 90% (18/20), based on normal productions (13 pairs) and defective articulation (5 pairs), and two defective articulation in different pairs.4. Speech profiling was performed based on perceptual and acoustic features of speech samples quasi-randomly selected from adult and child speech (65 samples; 58 speakers). Speaking fundamental frequency (i.e., vocal pitch level), speech rate, vocal tract length estimated from higher formant frequencies were key features which effectively explained individual differences of speech. Pitch range (variability), disfluency, and other perceptual features can be used to describe and identify individual speakers. Less