项目名称: 基于面部肌肉运动模型的三维人脸表情识别研究
项目编号: No.61202237
项目类型: 青年科学基金项目
立项/批准年度: 2013
项目学科: 计算机科学学科
项目作者: 黄迪
作者单位: 北京航空航天大学
项目金额: 25万元
中文摘要: 表情产生的本质是面部特定肌肉区域的运动,这种运动直接反映于人脸的三维几何信息的改变,所以基于三维数据的方法很可能满足对细微表情变化的完整描述。本项目结合表情产生的特点,拟提出基于面部肌肉运动模型的三维人脸表情识别方法,对产生表情的主要面部肌肉进行准确分割和独立建模,利用在这些面部肌肉变化过程中体现在人脸曲面上的几何差异进行表情区分,包含自动标志点定位,面部区域分割,特征提取以及分类器设计等研究内容,拟建立较为完善的三维人脸表情理解的理论框架,并在面部肌肉区域分割和层次化的几何特征表示方面做出独特性工作,并在理论研究的基础上,给出一些对于实际应用具有指导意义的结果,搭建演示系统,实现快速精确的三维人脸表情识别。
中文关键词: 三维人脸;标志点定位;曲面分割;曲面表示;表情识别
英文摘要: Facial expressions are caused due to the motion of muscles in specific face areas, which can be directly captured by the geometrical information of 3D face; therefore, the facial expression recognition approaches based on 3D data possess good potential to accurately and comprehensively represent subtle expression changes. Considering the characteristic of facial expressions, this project is going to propose a muscle motion based 3D facial expression recognition, including several important academic problems, i.e. automatic landmarking,facial region segmentation, geometric feature extraction and classifer design, aiming to construct a theoretical and systemetical framework for automatic 3D facial expression recognition especially highlight the nolvelty in facial muscle area segmentation and extraction of multi-level geometirc features. On the other side, we are going to achieve some practical conclusions and a demo system .
英文关键词: 3D Face;Landmarking;Surface Segmentation;Surface Representation;Expression Recognition