Development of computer-aided speech evaluation system modified by the technology of speech recognition

利用语音识别技术改进的计算机辅助语音评价系统的开发

基本信息

批准号：
10557193
负责人：
SUZUKI Noriko
金额：
$ 8.64万
依托单位：
Showa University
依托单位国家：
日本
项目类别：
Grant-in-Aid for Scientific Research (B).
财政年份：
1998
资助国家：
日本
起止时间：
1998 至 2000
项目状态：
已结题

项目摘要

The purpose of this research is to develope the computer-aided speech evaluation system which is modified by the technology of the speech recognition sysytem. For this purpose objective categorization of the distorted sound is neccesary to determine. Prior to this acoustical analysis, articulatory movement during abnormal sound production have to be analysed. During this period normal subjects and articulation disorders after head and neck cancer surgery are investigated.Following results were obtained :1-1. Three dimentional measurements of vocal tract shape using MRI(1) Parameter in the normal range exibited 1) stable change 2) the narrowest portion was centrally located on the palate.(2) The patlents vocal tract have markedwide and asymmetry caused by the surgery.1-2. FEM analyses of Three-Dimentional vocal tract modelsVocal transfer functions of three-dimensional vocal tract models estimated by a finite element method was shown. As a results, this method is more useful than one-dimentional models.2. Simultaneous analysis of mandibular and tongue movementA sirognathograph analysing system and an electropalatograph were used to record simultaneously. As a result, in the adults with moderate ankyloglossla, articulation of some sounds differed from that in the normal adults.3. Perception of Lateral Misarticulation (LM) and its physical correlatesAcoustical characteristics of LM were correlated to perceptional results.From these series of research, it was found that automatic machine recognition by computer is available in future.

这项研究的目的是开发计算机辅助语音评估系统，该系统通过语音识别系统的技术修改。为此，扭曲的声音的客观分类是必要的。在进行声学分析之前，必须分析异常声音产生期间的关节运动。在此期间，研究了头颈癌手术后的正常受试者和关节性疾病。取得良好的结果：1-1。使用MRI（1）参数在正常范围内进行的三维测量值1）稳定的变化2）最狭窄的部分位于口感上。（2）专利权人声道的声带标有遍及范围，并由手术引起的不对称。1-1-2。显示了对通过有限元方法估计的三维声道模型的三维声道模型的VEM分析。结果，此方法比一维模型更有用2。同时分析下颌和舌头运动sirageghogragghograph仪分析系统和电视造影仪同时记录记录。结果，在具有中度弧核的成年人中，某些声音的表达与正常成年人的声音不同。3。侧向错综复杂（LM）及其LM的物理相关性特征的感知与感知结果相关。从这些系列研究中，发现将来可以通过计算机识别自动机器识别。