基于发音运动分析与合成系统研究

发布时间：2018-06-11 19:51

本文选题：超声图像 + 发音可视化　；参考：《天津大学》2014年硕士论文

【摘要】：发音运动可视化是指将人的发音过程通过三维模型模拟出来。发音运动分析与合成系统涉及生理语音学、语音生理图像、人的发音机制及控制的生理计算模型等领域的相关技术,主要方法是建立超声数据库与三维生理模型数据库并将在二者之间建立映射关系从而实现发音运动的可视化,建立一个完善的发音运动分析与合成系统。目前,在对中文发音可视化研究方面存在欠缺,尤其是通过三维模型的可视化方面,因此发音运动分析与合成系统在语言学习,特别是第二语言学习及言语矫正方面以及言语矫正方面有很重要的意义。近年来,随着我们国家国际地位的提升,“汉语热”在很多地方兴起,现在汉语学习大部分都依靠“孔子学院”之类的传统教学培训机构,但是建立以及运营这些培训学校所需的人力、物资的消耗是非常巨大的。本文中我们采集了7个汉语原因因素,建立了超声数据库。在三维生理模型方面,为了涵盖大部分的中文发音器官的形态变化,我们计算了近20万组运动参数,通过大量计算建立了三维模型数据库。通过天津大学超算中心建立三维发声模型的数据库,以TPS(薄板样条函数)的方法实现超声图像到三维模型的映射,实现通过超声图像驱动三维模型,使发音运动可视化。最后,搭建了三维模型可视化系统,将于超声图像匹配后的三维模型进行输出,达到发音可视化的目的。
[Abstract]:The visualization of pronunciation movement is to simulate the process of pronunciation by three-dimensional model. The phonetic motion analysis and synthesis system involves related techniques in the fields of physiological phonetics, phonetics and images, human pronunciation mechanisms and controlled physiological computing models, etc. The main method is to establish the ultrasonic database and the three-dimensional physiological model database and establish the mapping relationship between them to realize the visualization of pronunciation movement and to establish a perfect phonetic motion analysis and synthesis system. At present, there are some deficiencies in the research of Chinese pronunciation visualization, especially through the visualization of 3D models, so the speech motion analysis and synthesis system is used in language learning. In particular, second language learning and speech correction are of great significance. In recent years, with the promotion of our country's international status, "Chinese craze" has risen in many places. Nowadays, most of our Chinese learning depends on traditional teaching and training institutions such as Confucius Institutes. But the human and material costs of building and running these schools are enormous. In this paper, we collected 7 Chinese cause factors and established the ultrasonic database. In order to cover most of the morphological changes of Chinese phonetic organs, nearly 200000 groups of motion parameters were calculated and the 3D model database was established by a large number of calculations. In this paper, the database of 3D phonation model is established in Tianjin University supercomputer Center, and the mapping from ultrasonic image to 3D model is realized by using TPS (thin sheet spline function) method. The 3D model is driven by ultrasonic image and the pronunciation motion is visualized. Finally, a 3D model visualization system is built to output the 3D model after ultrasonic image matching to achieve the purpose of pronunciation visualization.
【学位授予单位】：天津大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN912.3;TB559

【参考文献】