摘要
本研究着眼于汉语"近距离区域方音"比较与区分,比较元音和声调在三个近距离区域之间的差异,并通过线性判别模型建模,分析元音和声调相应的声学特征变量在自动区分三个区域中的作用。本实验共涉及元音第一、第二共振峰、声调平均基频、声调时长、声调三个极值点、声调目标相对位置,共九个变量,成功证明汉语近距离区域方音的部分特征变量的声学数据存在显著性差异。而后,本研究采取线性判别分析对现有数据进行建模,得出两个有效区分率高于85%的模型,即"平均基频+声调时长+元音F2"顺位加入模型和"平均基频+声调时长+极值点1基频值"顺位加入模型。线性判别分析模型的成功证明了汉语近距离方音采取数字模型进行区分的可行性,同时也证明了声调的"平均基频""基频时长"和"极值点"声学特征在汉语方音区分中发挥着重要作用。本研究也证实了"三点极值模型"能有效地、维度地测量声调。
The present study investigates the role of vocalic and tonal features as acoustic cues for geographically close sub-dialect identification by targeting three subregional dialects of the Xinyang dialect from China,as a first exploration of subregional dialects in languages with lexical tones.This study comprises two parts:an acoustic comparison of 9 correlates of the four lexical tones and three corner vowels in the three sub-dialects;a Linear Discriminant Analysis using all above acoustic correlates to identify the best acoustical predictor(s)for identifying the three regional dialects.The present study showed that three sub-dialects differ significantly with respect to the corner vowels and the four lexical tones.These differences,including mean pitch,pitch duration and pivot value,are effective enough to be successful in statistical acoustic classification.
出处
《语言研究集刊》
2020年第2期342-362,445-446,共23页
Bulletin of Linguistic Studies
