Comparison of PCA, LDA and Gabor Features for Face Recognition Using Fuzzy Neural Network

摘要

A face recognition system identifies or verifies face images from a stored database of faces when a still image or a video is given as input. The recognition accuracy depends on the features used to represent the face images. In this paper a comparison of three popular features - PCA, LDA and Gabor features-used in literature to represent face images is given. The classifier used is a Fuzzy Neural Network classifier. The comparison was performed using AT&T, Yale and Indian databases. From the experimental results, the LDA features provide better Recognition Rates in the case of face images with less pose variations. Where more pose variations are involved, the Gabor features performed better than LDA features. For recognition tasks where recognition of trained individuals and rejection of untrained individuals are considered, the LDA features provide better results in terms of very low False Acceptance Rates and False Rejection Rates.