Readers 读者
A two-dimensional lip model for Mandarin Chinese
汉语普通话二维唇模型
Xiaosheng Pan 潘晓声
Abstract 摘要
A two-dimensional lip model with inner lip information is provided in this paper.1 The lip model can precisely describe all possible lip states, which generally include complete closure, half opening, and full opening. By changing the values of the lip model’s parameters, all kinds of lip contours with different shapes can be generated, and the lip contours can fit the real lip shape quite well. The lip geometrical features based on lip contour are extracted and used to define a basic linguistic term, lip rounding. This paper has shown that the width of the inner lip is as important as the width of the outer lip when defining lip rounding. In the Audio-Visual (AV) speech perception experiment of the McGurk Effect, samples are developed with the aid of the lip model. The first sample is to keep all lip information including the tongue and the teeth information, the second sample is to keep both outer and inner lip information, and the last sample is to keep the outer lip information only. The result proves that the inner lip provides the most visual information of speech in the McGurk Effect experiment. It is concluded that the inner lip information is indispensable in the lip model for linguistic study based on the above research.
本文提出了一种具用内唇信息的二维唇模型。此模型可以精确的描述所有可能的唇形状态，包括全闭合状态，半开合状态和全张开状态。通过调整模型的参数，可以生成各种不同形状的唇形，并能完全拟合真实嘴唇轮廓。本文提取了基于嘴唇轮廓的唇形几何特征，并用之定义语言学的一个基本概念，圆唇。本文的研究结果表明在定义圆唇时，内唇宽度和外唇宽度的重要性相似。利用本文提出的唇模型所生成测试样本，将其用于McGurk效应的可视语音感知实验中。第一个样本保留了外唇轮廓信息，第二个样本同时保留了内外唇轮廓信息，最后一个样本除了保留内外唇轮廓信息之外还提供了舌头和牙齿的信息。实验结果证明在McGurk效应的实验中内唇轮廓提供了语言的大部份的视觉信息。可以由此得出在涉及语言的研究中，唇模型的内唇信息是必不可少的。
Keywords 关键词
Lip model 唇模型 Lip rounding 圆唇 Audio-Visual speech perception 可视语音感知 McGurk Effect McGurk 效应