INTELLIGENT FACE RECOGNITION SYSTEMS

Abstract

Face recognition methods and algorithms have been improved during the last years. A lot of research and studies have been done to establish high accuracy and fast recognition rate in face recognition systems. Although various results were estimated using different techniques to reach best accuracy and performance. This leads us to continue the wheel of improvements to conduct more studies about face recognition techniques. In this thesis we make comparison with the most known traditional technique of face recognition EigenFace using principal component analysis (PCA) algorithm, Linear discriminant analysis (LDA) Fisher face approach and Local Binary Patterns (LBP). An enhanced comparison with some of the most recent advanced techniques related to deep learning and neural networks. Results shows that advanced techniques that depend on deep learning algorithms outperform traditional techniques in terms of accuracy and computational time. On the other hand, among the traditional tested techniques, we notice that LBP gives the best accuracy with 96% and 89% when compared using the CALTECH and FEI datasets respectively