A Face Recognition System using Neural Networks with Incremental Learning Ability

In this paper, we present a new incremental face recognition (IFR) system based on new adaptive learning algorithms and networks. We introduce new adaptive linear discriminant analysis (LDA) algorithm and related network for optimal facial feature extraction and use them to construct a new IFR system. Convergence proof of all algorithms is given using an appropriate cost function and discussing about its initial conditions. Application of the new IFR on feature extraction from facial image sequences is given in two steps: i) image preprocessing, which includes normalization, histogram equalization, mean centering and background removal, ii) adaptive LDA feature estimation. In the preprocessing stage, all input images are cropped and prepared for the next step. Outputs of the preprocessing stage are used as a sequence of inputs for IFR system. The proposed system was tested on YALE face database. Experimental results on this database demonstrated the effectiveness of the proposed system for adaptive estimation of the feature space for online face recognition.