A novel Fisher discriminant for biometrics recognition: 2DPCA plus 2DFLD

— this paper presents a novel image feature extraction and recognition method two dimensional linear discriminant analysis (2DLDA) in a much smaller subspace. Image representation and recognition based on the Fisher’s criterion is statistically dependent on the evaluation of the covariance matrices. Since the proposed approach computes the covariance matrices in a subspace of the input space, the optimal discriminant vectors are more accurately obtained. Furthermore, the proposed method is based on 2D image matrices rather than 1D vector so the image does not need to be transformed into a vector prior to feature extraction. This leads to the following benefits; the proposed method yields greater recognition accuracy while reduces the overall computational complexity. Finally, the effectiveness of the proposed algorithm is verified using the ORL database as a benchmark. The new algorithm achieves a recognition rate of 95.50% compared to the recognition rate of 90.00% for the Fisherf...

• Type: Conference
• Year: 2006
• Where: ISCAS
• Authors: R. M. Mutelo, Li Chin Khor, Wai Lok Woo, Satnam Singh Dlay
• Keywords: Covariance Matrices, Feature Extraction, Image Feature Extraction, ISCAS 2006