Open AccessSparse Group Regression Classification for Face Recognition
Author(s) -
Zheng Cheng-yong,
Zhi-yu Luo
Publication year - 2021
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1815/1/012017
Subject(s) - regression , pattern recognition (psychology) , facial recognition system , sparse approximation , face (sociological concept) , artificial intelligence , regression analysis , computer science , biometrics , mathematics , statistics , machine learning , social science , sociology
Face recognition has been widely used in biometric verification which is a significant issue in system control in computer based communication. Sparse regression classification (SRC) has been an important method for face recognition. SRC assumes its regression coefficients are globally sparse. However, these regression coefficients correspond to a class-specific gallery may be locally dense. In order to get some regression coefficients that are globally sparse and locally dense, sparse group regression (SGR) is introduced in this paper for face recognition. By use of SGR, we expect to get some regression coefficients which are not only sparse as a whole, but also locally dense in some part. Variable splitting and alternating direction method of multipliers (ADMM) are used for solving the problem of SGR. The proposed algorithm, named Sparse Group Regression Classification (SGRC), is extensively evaluated on four standard databases. A comparative study with some state-of-the-art algorithms reflects the efficacy of the proposed approach.