Linear subspaces for facial expression recognition
作者：
Highlights：
• Vectors formed by facial landmark point coordinates are used.
• We assume that these vectors belong to a different manifold for each of the expressions.
• Expression sequences are represented by their principal direction.
• Manifolds are decomposed to a small number of linear subspaces of low dimension.
• Subject-independent and subject-dependent experiments are performed.
摘要
Highlights•Vectors formed by facial landmark point coordinates are used.•We assume that these vectors belong to a different manifold for each of the expressions.•Expression sequences are represented by their principal direction.•Manifolds are decomposed to a small number of linear subspaces of low dimension.•Subject-independent and subject-dependent experiments are performed.
论文关键词：Face analysis,Expression recognition,Subspaces
论文评审过程：Received 25 March 2012, Revised 26 May 2013, Accepted 22 October 2013, Available online 5 November 2013.
论文官网地址：https://doi.org/10.1016/j.image.2013.10.004
