Linear Discriminant Classifier Ensemble for Face Recognition

摘要

Face recognition is an important field of research in a variety of domains including law enforcement, commerce, social media and marketing. We have developed a Linear Discriminant Analysis (LDA) based Ensemble Classification system for face recognition purposes. The proposed ensemble uses multi-resolution Wavelet analysis to extract features from two benchmark databases i.e. the ORL (AT&T) and the Yale image databases. Principal Component Analysis (PCA) has been employed for dimensionality reduction. The multi-resolution features are combined for efficient utilization for classification purposes. The ensemble has been trained using the Bagging technique and cross validated to obtain classification accuracy of 98.27% for the ORL database and 95.15% for the Yale database.