Abstract

Face recognition has been a very active research area for past two decades due to its widely applications such as identity authentication, airport security and access control, surveillance, and video retrieval systems, etc. Numerous approaches have been proposed for face recognition and considerable successes have been reported [1]. A successful face recognition system should be robust under a variety of conditions, such as varying illuminations, pose, expression, and backgrounds. Although many researches [1-3] already found solutions to alleviate those problems, it is still a great challenge to accurately recognize faces that are non-frontal facial images, and disguises such as facial hair, glasses or cosmetics. In the past, two categories of feature extraction for face recognition have been developed, including of geometric features and statistical features derived from face images [1-4]. In geometric features method, the facial features are retrieved from either the shapes of eyes, nose, mouth, and chin, or the facial geometrical relationships such as areas, distances, and angles [4]. This kind of approaches has proven to be difficult for practical application because it requires a fine segmentation of facial features. In statistical features method, the facial features are usually extracted from important facial features based on the highdimensional intensity values of face images. For example, the principal component analysis (PCA) is the well-known statistical method [5]. This approach is simple, but does not reflect the details of facial local features. Generally speaking, researches on face recognition system can be grouped into two categories of classifier system, one is single-classifier system and the other is multi-classifier system. The single-classifier systems, including neural network (NN) [6], Eigenface [5], Fisher linear discriminant (FLD) [7], SVM [8], HMM [9], and AdaBoost [10], are developed well in theory and experiment. On the other hand, the multi-classifier systems such as local and global face information fusion [11-13], neural networks committee (NNC) [14], multiclassifier system (MCS) [15], are proposed in parallel process of different features or classifiers. As one knows, neural networks (NN) are a nonlinear classifier and based on the parallel architecture of human brains. Its output can be binary or multiple classes. In NN, the margin between two classes of sample point nearby the decision function is often not a maximum. O pe n A cc es s D at ab as e w w w .in te ch w eb .o rg

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call