Far East Journal of Experimental and Theoretical Artificial Intelligence
Volume 4, Issue 2, Pages 89 - 101
(November 2009)
|
|
FACE RECOGNITION BASED ON MULTI-CLASS SUPPORT VECTOR MACHINE AND FISHER LINEAR DISCRIMINANT METHOD
Deng-Yuan Huang (Taiwan), Wu-Chih Hu (Taiwan), Chun-Jih Lin (Taiwan) and Mu-Song Chen (Taiwan)
|
Abstract:
In this paper, we present an efficient method of LDA+SVM with two-scale Haar wavelet transformation for face recognition. The performance is evaluated by the ORL and MIT-CBCL databases. The dimensionality reduction of face feature spaces is first carried out by the PCA, LDA and D-LDA methods to form a discriminative feature vector. The face classifications are then performed by the methods of support vector machine (SVM), Euclidean distance, and cosine distance. The recognition accuracies of LDA+SVM achieve 96% and 97% for the evaluations of the ORL and MIT-CBCL databases, respectively. This result also shows the excellence of the proposed method. |
| Keywords and phrases: face recognition, PCA, LDA, SVM. |
| Communicated by Chuan-Yu Chang |
| Number of Downloads: 149 | Number of Views: 501 |
|
