语音接口在汉语学习寓教于乐系统中的应用

发布时间：2018-04-22 03:21

本文选题：语音识别 + 语音评测　；参考：《北京交通大学》2009年硕士论文

【摘要】： 近些年来,随着中国经济的快速发展和中国国际地位的不断提高,中国与世界的交往和联系日趋广泛和深入。汉语是中华文化的主要载体,也是世界各国了解中国的重要工具,不少国家出现了学习汉语的热潮。但是在全球汉语学习迅速升温的同时也带来了一些问题,如汉语教学资源不足和传统教学方式不能有效地激发学生学习汉语的兴趣等。而寓教于乐的学习形式能够很好地解决这些问题。自2004年在美国首先提出了这个概念后,它已经在众多领域取得了丰硕的成果,但目前只有少数的研究者致力于外国汉语学习者的寓教于乐教学研究。针对现在对外汉语教学中出现的问题,我们提出了一种基于语音接口的汉语学习寓教于乐系统,以方便学生自我学习并提高汉语学习的兴趣。本论文的主要工作如下: (1)利用HTK平台建立了一个非特定人孤立词语音识别系统,并从混合高斯模型、语言模型和基频参数等方面对该系统进行改进,最终把系统的识别率提高到98%以上,基本满足了实际的使用要求。 (2)改进了HTK的识别器HVITE,使它能够输出字词和声韵母两个层次的识别结果信息和发音评测结果,为寓教于乐系统的语音接口做好后台处理准备。 (3)提出了一种新的基于HMM的对数似然值与声韵母层时长信息相结合的发音评测方法。本评分方法对专家评分的相关度高于基于HMM的后验概率的方法。通过求解非线性回归模型和模型参数优化,建立两个统一的声韵母对数似然值的映射模型;并将最终的评分及映射模型嵌入HTK的识别器HVITE中。 (4)利用Virtools软件平台,设计并实现汉语发音练习的寓教于乐系统。本系统主要包括建立虚拟现实场景和各个角色模型及其相关的动作,并且将实验室三维重建的成果成功的应用到系统中;利用Virtools中的SDK开发包,在VC++6.0平台上开发适用于本系统的语音识别与评测接口,使本系统实现纠正学生声韵母层汉语发音的功能。
[Abstract]:In recent years, with the rapid development of China's economy and the continuous improvement of China's international status, China's contacts and contacts with the world are increasingly extensive and in-depth. Chinese is not only the main carrier of Chinese culture, but also an important tool for all countries to understand China. However, with the rapid increase of global Chinese learning, some problems have been brought, such as the shortage of Chinese teaching resources and the inability of traditional teaching methods to stimulate students' interest in learning Chinese effectively. And the form of learning with pleasure can solve these problems very well. Since it was first proposed in the United States in 2004, it has achieved fruitful results in many fields, but at present only a few researchers are devoted to the teaching and learning of foreign Chinese learners. In view of the problems in teaching Chinese as a foreign language, we propose a Chinese learning system based on phonetic interface, which can facilitate students' self-learning and improve their interest in learning Chinese. The main work of this thesis is as follows: 1) using HTK platform, a speech recognition system for isolated words is established, and the system is improved from the aspects of mixed Gao Si model, language model and fundamental frequency parameters. Finally, the recognition rate of the system is increased to more than 98%. Basically met the actual use requirements. (2) the HTK recognizer HVITE is improved to output the recognition result information and pronunciation evaluation result at the two levels of word and vowel, so as to prepare the background processing for the phonetic interface of the teaching music system. A new pronunciation evaluation method based on HMM is proposed. The relevance of this method to expert score is higher than that of posterior probability method based on HMM. By solving nonlinear regression model and model parameter optimization, two unified mapping models of logarithmic likelihood value of rhyme and mother are established, and the final score and mapping model are embedded in the recognizer HVITE of HTK. Using Virtools software platform, we design and implement the Chinese pronunciation practice system. The system mainly includes the establishment of virtual reality scene and each role model and its related actions, and the successful application of the results of 3D reconstruction in the laboratory to the system, and the use of SDK development kit in Virtools. The interface of speech recognition and evaluation is developed on the platform of VC 6.0, which makes the system realize the function of correcting students' phonetic master level Chinese pronunciation.
【学位授予单位】：北京交通大学
【学位级别】：硕士
【学位授予年份】：2009
【分类号】：TN912.34

【相似文献】