Face recognition based on symmetrical weighted PCA

This paper presents a novel symmetrical weighted principal component analysis (SWPCA) space for feature extraction and its application to face recognition. Specifically, SWPCA first applies mirror transform to facial images, and gets the odd and even symmetrical images based on the odd-even decomposition theory. Then, weighted PCA is performed on the odd and even symmetrical training sample sets respectively to extract facial image features. Finally, nearest neighbor classifier is employed for classification. SWPCA method was tested on face recognition using the ORL, Yale and FERET databases, where the images vary in illumination, facial expression, poses and scale. SWPCA achieves 96% correct face recognition rate for ORL database, 97.778% accuracy for Yale database and 96.19% accuracy for FERET database. Experiments also demonstrate that SWPCA has better recognition accuracy comparing with conventional approaches such as PCA, SPCA and WPCA.