DocumentCode :
596489
Title :
Real-time facial landmarks tracking using active shape model and LK optical flow
Author :
Byungtae Ahn ; Yudeog Han ; In So Kweon
Author_Institution :
Korea Adv. Inst. of Sci. & Technol., Daejeon, South Korea
fYear :
2012
fDate :
26-28 Nov. 2012
Firstpage :
541
Lastpage :
543
Abstract :
Active Shape Models (ASM) is a generative model widely used to model faces. ASM has been successfully used for face and emotion recognition, and it is one of the state-of-the-art approaches because of its efficiency and representational power. Although widely employed, applying only ASM is not adequate for the practical applications, because positions of the facial landmarks are unstably extracted like jittering movements in the sequential frames, which degrades the performance of the applications. In this paper, we propose a framework for real-time facial landmarks extraction and tracking using ASM and Lucas-Kanade (LK) optical flow which is considered desirable to estimate time-varying geometric parameters in sequential dynamic images of face. In addition, we introduce a straightforward method to avoid failure to extract the facial landmarks by occlusion using the positions of the extracted landmarks by ASM and tracked by LK optical flow. Experimental results validate our approach.
Keywords :
computational geometry; emotion recognition; face recognition; image sequences; object tracking; ASM; LK optical flow; Lucas-Kanade optical flow; active shape model; emotion recognition; face recognition; generative model; jittering movements; real-time facial landmarks tracking; sequential dynamic images; time-varying geometric parameters; Ambient intelligence; Robots; ASM; LK optical flow; MRASM; occlusion;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Ubiquitous Robots and Ambient Intelligence (URAI), 2012 9th International Conference on
Conference_Location :
Daejeon
Print_ISBN :
978-1-4673-3111-1
Electronic_ISBN :
978-1-4673-3110-4
Type :
conf
DOI :
10.1109/URAI.2012.6463068
Filename :
6463068
Link To Document :
بازگشت