Quadratic projection based feature extraction with its application to biometric recognition
作者：
Highlights：
• A novel quadratic projection based feature extraction framework is developed.
• A set of quadratic matrices is learnt to distinguish each class from other classes.
• Quadratic matrix learning (QML) is formulated as an SDP problem.
• An efficient algorithm is developed to solve QML based on Lagrange duality theory.
• Experiments on biometric recognition show the effectiveness of our algorithm.
摘要
Highlights•A novel quadratic projection based feature extraction framework is developed.•A set of quadratic matrices is learnt to distinguish each class from other classes.•Quadratic matrix learning (QML) is formulated as an SDP problem.•An efficient algorithm is developed to solve QML based on Lagrange duality theory.•Experiments on biometric recognition show the effectiveness of our algorithm.
论文关键词：Biometric recognition,Feature extraction,Quadratic projection,Semidefinite programming,Lagrange duality
论文评审过程：Received 16 July 2015, Revised 7 December 2015, Accepted 16 February 2016, Available online 24 February 2016, Version of Record 12 April 2016.
论文官网地址：https://doi.org/10.1016/j.patcog.2016.02.010
