QQ客服 官方微博 反馈留言

	Automatic Facial Expression Recognition on a Single 3D Face by Exploring Shape Deformation
	Boqing Gong; Yueming Wang; Jianzhuang Liu; Xiaoou Tang
	2009
会议名称	17th ACM International Conference on Multimedia, MM'09, with Co-located Workshops and Symposiums
英文摘要	Facial expression recognition has many applications in multimedia processing and the development of 3D data acquisition techniques makes it possible to identify expressions using 3D shape information. In this paper, we propose an automatic facial expression recognition approach based on a single 3D face. The shape of an expressional 3D face is approximated as the sum of two parts, a basic facial shape component (BFSC) and an expressional shape component (ESC). The BFSC represents the basic face structure and neutral-style shape and the ESC contains shape changes caused by facial expressions. To separate the BFSC and ESC, our method firstly builds a reference face for each input3D non-neutral face by a learning method, which well represents the basic facial shape. Then, based on the BFSC and the original expressionalface, a facial expression descriptor is designed. The surface depth changes are considered in the descriptor. Finally, the descriptor is input into an SVM to recognize the expression. Unlike previous methods which recognize a facial expression with the help of manually labeled key points and/or a neutral face, our method works on a single 3D face without any manual assistance. Extensive experiments are carried out on the BU-3DFE database and comparisons with existing methods are conducted. The experimental results show the effectiveness of our method.
收录类别	EI
语种	英语
内容类型	会议论文
源URL	[http://ir.siat.ac.cn:8080/handle/172644/2387]
专题	深圳先进技术研究院_集成所
作者单位	2009
推荐引用方式 GB/T 7714	Boqing Gong,Yueming Wang,Jianzhuang Liu,et al. Automatic Facial Expression Recognition on a Single 3D Face by Exploring Shape Deformation[C]. 见:17th ACM International Conference on Multimedia, MM'09, with Co-located Workshops and Symposiums.

个性服务

查看访问统计

相关权益政策

暂无数据

收藏/分享

除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。