Title :
Face recognition based on Multi-class Fisher Scores
Author :
Haiying, Xu ; Jiao, He ; Chaocheng, Xie ; Xicheng Liu
Author_Institution :
Sch. of Electron. & Inf. Eng., Southwest Pet. Univ., Cheng Du, China
Abstract :
In the use of hidden Markov model for face recognition, discrimination is weak, and the parameters of the model require a higher degree of precision. Therefore, this paper proposes a new method to obtain Fisher score features. This has fully considered the model parameters on the categories of the differential contribution, so precision of hidden Markov model parameters is lower. Because sole class discrimination is limited, this paper attempts to use multi-class Fisher score feature series in order to further improve the characteristics of the type of discrimination, the experiments proved that the Fisher score characteristics can greatly improve the face recognition rate.
Keywords :
face recognition; feature extraction; hidden Markov models; matrix algebra; Fisher kernel; Fisher score characteristics; Fisher score feature; Riemann manifold; class discrimination; face recognition; hidden Markov model; information matrix; kernel transformation; model parameters; multiclass Fisher score; Face; Face recognition; Hidden Markov models; Kernel; Support vector machine classification; Training; Vectors; Fisher discrimination analysis; Fisher scores; face recognition; multi-class mapping;
Conference_Titel :
Fuzzy Systems and Knowledge Discovery (FSKD), 2012 9th International Conference on
Conference_Location :
Sichuan
Print_ISBN :
978-1-4673-0025-4
DOI :
10.1109/FSKD.2012.6234309