权益分类	功能权益	普通用户	{{item.name}}会员
{{category.name}}	{{benefitItem.name}}

Research on recognition and synthetic system of individual characteristic in talk and face expression

谈话与面部表情个体特征识别与合成系统研究

基本信息

批准号：
11680402
负责人：
OGAWA Hitoshi
金额：
$ 2.43万
依托单位：
Ritsumeikan University
依托单位国家：
日本
项目类别：
Grant-in-Aid for Scientific Research (C)
财政年份：
1999
资助国家：
日本
起止时间：
1999 至 2001
项目状态：
已结题

项目摘要

To do the voice synthesis of a human spoken language which is corresponded to a natural lace expression and the face as a basic technology necessary for achieving the personifying agent who suited the human interlace, the following research was done.1. The three-dimensional face model to achieve a natural face expression was made. The face expression system was developed by using FACS (Facial Action Coding System). The shape of the mouth in a Japanese utterance was classified to achieve natural rip synchronization matched to a synthetic sound efficiently, and it was confirmed that it was enough in the shape of nine kinds of mouths.2. To develop the technique to control a basic frequency by language information and non-language information, and to examine the basic frequency pattern of the utterance, the following two methods were examined.(1) It turned out to be able to achieve a unique utterance in a basic frequency of everybody because a basic frequency of the utterance was able to be reverse-converted.(2) The measurement method was proposed by using the idea of the break-up phrase of the change pattern in sentences of a basic frequency, and effectiveness was confirmed by the experiment.3. The voice synthesis system and the face dynamic scene generation system of a basic theory because of the delicious synchronization were researched. The automatic study method of the strategy knowledge was researched to hold a dynamic correction technique of the scheduling technique and the schedule which considered time and efficient plans, and the result was achieved.

为了将与自然表情和面部相对应的人类口语的语音合成作为实现适合人类界面的拟人化剂所必需的基本技术，进行了以下研究.建立了三维人脸模型，实现了自然的人脸表情。人脸表情系统是利用人脸动作编码系统（FACS）开发的。对日语发音中的嘴的形状进行分类，以有效地实现与合成声音匹配的自然裂口同步，并确认在9种嘴的形状中就足够了。为了开发通过语言信息和非语言信息来控制基本频率的技术，并且为了检查话语的基本频率模式，检查了以下两种方法。(1)它能够在每个人的基本频率中实现唯一的话语，因为话语的基本频率能够被反向转换。(2)提出了基于基频句中变化模式分解短语的测量方法，并通过实验验证了该方法的有效性.研究了语音合成系统和人脸动态场景生成系统的一个基本理论，因为美味的同步。研究了策略知识的自动学习方法，实现了调度技术的动态修正技术和兼顾时间和效率计划的调度方法，并取得了较好的效果。