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

A Common Platform for Analysis and Synthesis of Face to Realize a Kansei Communication

实现感性交流的人脸分析与合成通用平台

基本信息

批准号：
11555097
负责人：
HARASHIMA Hiroshi
金额：
$ 8.19万
依托单位：
The University of Tokyo
依托单位国家：
日本
项目类别：
Grant-in-Aid for Scientific Research (B)
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

来源：
https://kaken.nii.ac.jp/en/grant/KAKENHI-PROJECT-11555097/
关键词：
Face recognition Face image synthesis Face model fitting Common software Face animation Kansei communication Life-like agent Face-to-face interface 顔分析表情合成平均顔感性処理 3次元モデル顔情報処理顔合成特徴抽出表情変形 GUI

项目摘要

Until now, several research institutes and university have tried a face recognition and face image synthesis schemes and they just have come to produce a commercial product. However, most of the elemental technologies are realized about case by case requirement and they don't have a common interface with other technologies and extension function. In this background, our project tried to realize to construct a common software system about face image processing with generalized hierarchical function. The major element of this software system is composed of a common face processing library which can treat both face recognition and face synthesis. In the face recognition system, an extraction and parameterization of face and face parts, tracking of face in video sequence, basic tool for person identification and expression recognition are included. And also API is produced to make generalization of system. In the face image synthesis system, 3D model and facial expression rule to generate a natural and real face image and animation are constructed. A generic face model is modified to be fitted automatically to face frontal picture based on feature points extraction result by face recognition system to make customization of face model. One new point of our research is a combination of each algorithm produced independently by several research institutes to make generalization of system. A major improvement of animation quality of facial expression is second point. As an interaction technology, a treatment of emotion condition and communication through narrow band network is also considered, and common markup language to realize a communication on WEB page is also proposed as a new result. By psychological analysis, facial expression animation rule on time scale is exactly treated.

到目前为止，一些研究机构和大学已经尝试了人脸识别和人脸图像合成方案，他们刚刚开始生产商业产品。然而，大多数基本技术都是针对具体需求实现的，它们与其他技术没有一个通用的接口和扩展功能。在此背景下，本课题尝试构建一个具有广义层次函数的通用人脸图像处理软件系统。该软件系统的主要组成部分是一个通用的人脸处理库，可以处理人脸识别和人脸合成。在人脸识别系统中，包括人脸和人脸部分的提取和参数化，视频序列中人脸的跟踪，基本的身份识别和表情识别工具。为了使系统通用化，还设计了API。在人脸图像合成系统中，构建了三维模型和人脸表情规则，生成了自然的、真实的人脸图像和动画。根据人脸识别系统的特征点提取结果，对通用人脸模型进行修改，使其自动拟合到人脸正面图像，从而实现人脸模型的定制。本文的一个创新点是将多个研究机构各自独立提出的算法结合起来，使系统具有通用性。第二点主要是提高了动画的面部表情质量.作为一种交互技术，还考虑了通过窄带网络的情感状态和通信的处理，并且作为新的结果，还提出了在WEB页面上实现通信的公共标记语言。通过心理学分析，准确地处理了表情动画在时间尺度上的规律。