由于遗传、地域、文化等因素的不同，各民族在人脸情感表达上具有一定的差异性。本项目拟利用KINECT平台采集包含深度信息在内的人脸表情视频， 构建中国多民族人脸表情库并进行发布，通过提出族群谱系特征并融合人脸深度信息对民族情感表达进行深入分析和探讨；扩展公理化模糊集（AFS）体系构建人脸族群谱系特征分析方法，基于FACS编码体系结合迁移学习构建分类器对多民族表情库进行标识和知识挖掘；通过多民族图模式情感AU关联模型对族群表情进行知识表达，并研究相应的图模式比较和模式挖掘算法。据此探求多民族在基本表情、非基本和动态表情的共通性和差异性，发掘其人脸信息、表情与情感之间的内在规律。研究工作将三维数据、族谱特征和AU模式挖掘结合在一起以提高民族面部情感分析效率，也为相关研究提供可借鉴的手段，其成果对多民族情感数字化留存、民族表情生成和深化人类学民族特征研究也有一定的意义。

Different ethnic groups have different characteristics in the facial expressions, it dues to the difference of genetic, geography and culture. With KINECT, the project plans to collect the expression videos with facial depth information for constructing and publishing the multi-ethnic facial expression database, analyzes deeply the multi-ethnic emotion with ethnic groups spectrum characteristics and facial depth information. Firstly, the analysis of facial ethnic groups spectrum characteristics is constructed by Axiomatic Fuzzy Sets(AFS). Secondly, combining the FACS and transfer learning classifier, the classifier is established to create multi-ethnic facial expression database and discover ethnic groups knowledge. At last, the knowledge of ethnic groups expression is represented through AU association graph pattern model, meanwhile the corresponding graph pattern evaluation and pattern mining algorithm is studied. According to the above, this project explores the commonalities and differences of multi-ethnic expressions in basic, non-basic, and dynamic expression, also reveals facial inherent knowledge of information, expression and emotion. This project prepares to enhance efficiency of multi-ethnic facial emotion analysis, ethnic groups spectrum characteristics and AU graph pattern mining, also provide many methods to the relevant researches. Especially, this project has a great significance of digital storage of multi-ethnic emotion, generation of multi-ethnic expression and study of anthropology multi-ethnic characteristics.