Attention driven face recognition: A combination of spatial variant fixations and glance

作者：Wang Chongxiu^*; Qing Laiyun; Miao Jun; Fang Fang; Chen Xilin

来源：2011 IEEE International Conference on Automatic Face and Gesture Recognition and Workshops, FG 2011, 2011-03-21 to 2011-03-25.

DOI：10.1109/FG.2011.5771340

• Computational costs
• Face images
• Face recognition algorithms
• Face recognition systems
• FERET database
• Foveated imaging
• Human Visual System
• Human visual systems
• Low-spatial frequency
• Selective attention
• Spatial acuity
• Spatial-variant

This paper proposes a novel face recognition algorithm inspired by the selective attention of Human Visual System (HVS). We record four observers'; eye movements when they are viewing 100 FRGC [1] frontal view face images and find that the observers are highly consistent in the regions fixated. Inspired by the fact that fovea of HVS has a much higher spatial acuity than the periphery, a face recognition algorithm based on spatial variant sampling is proposed to simulate such foveated imaging phenomenon, where more information is reserved for the fixated regions. Moreover, information extracted from glance which adopts the low spatial frequency components of the image is integrated into the face recognition system to elicit a percept that occurs before any fixations. The experimental results on FERET database [2] demonstrate that the proposed method not only reduces the computational cost, but also achieves comparable performance, which shows that the characteristics of the HVS provide valuable insights into face recognition.

• 出版日期2011