中文摘要：

本文着眼于语音可视化,目的在于揭示中英文辅音的发音运动。然而,交叉语言比较的首要任务是消除特定说话人生理特性等音素引起的固有差别,本文采用了基于说话人归一化的普氏算法来进行说话人归一化处理,接着利用分层聚类分析算法和多维标度定位算法对相似发音对进行了量化的比较,从而获得了英汉相似音素发音的元音对比最小对和辅音对比最小对,并揭示了这2种语言的易混淆发音文本对在发音动作上的区别。

英文摘要：

This paper focuses on speech visualization, and its targets is designed for English-Mandarin consonants articulatory movement. However, the first task of cross linguistic comparison is to eliminate the variability which came from speaker-specific vocal-tract structure and the other individual biomechanical properties. This paper uses the speaker normalization based Procrustes to realize speaker normalization, and uses hierarchical clustering algorithm and multidimensional calibration algorithm to make quantitative comparison of similar pronunciations pairs, so as to obtain bilingual vowel and consonant mini-pairs respectively. Therefore, bilingual confusing pronunciation vowel text pairs differences of internal and external motions are revealed successfully.