DocumentCode :
234371
Title :
3D face recognition using facial curves, sparse random projection and fuzzy similarity measure
Author :
Belghini, Naouar ; Ezghari, Soufiane ; Zahi, Azeddine
Author_Institution :
Syst. Intell. & Applic. Lab. (SIA, FST, Fez, Morocco
fYear :
2014
fDate :
20-22 Oct. 2014
Firstpage :
317
Lastpage :
322
Abstract :
In this paper, we propose a fuzzy similarity based classification approach for 3D face recognition. In the feature extraction method, we exploit curve concept to represent the 3D facial data, two types of curves was considered: depth-level and depth-radial curves. As the dimension of the obtained features is high, the problem “curse of dimensionality” appears. To solve this problem, the Random Projection (RP) method was used. The proposed classifier performs Fuzzification operation using triangular membership functions for input data and ordered weighted averaging operators to measure similarity. Experiment was conducted using vrml files from 3D Database considering only one training sample per person. The obtained results are very promising for depth-level and depth-radial curves, besides the recognition rates are higher than 98%.
Keywords :
face recognition; feature extraction; fuzzy set theory; image classification; 3D database; 3D face recognition; 3D facial data; RP method; depth-level curves; depth-radial curves; dimensionality curse; facial curves; feature extraction method; fuzzification operation; fuzzy similarity based classification; ordered weighted averaging operators; random projection method; sparse random projection; triangular membership functions; vrml files; Abstracts; Decision support systems; Face recognition; Feature extraction; Knowledge based systems; Pragmatics; Three-dimensional displays; 3D face recognition; OWA operator; facial curves; fuzzy logic; similarity measure; sparse random projection;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Information Science and Technology (CIST), 2014 Third IEEE International Colloquium in
Conference_Location :
Tetouan
Print_ISBN :
978-1-4799-5978-5
Type :
conf
DOI :
10.1109/CIST.2014.7016639
Filename :
7016639
Link To Document :
بازگشت