Face recognition using AAM and global shape features

Abstract

A new technique for face recognition is proposed, which uses Active Appearance Model (AAM) to extract facial feature points and uses global shape features to recognize face. To enhance performance of AAM, we use Adaboost to locate positions of eyes. After extraction of facial feature points, we use any two points of global shape features and compute the distance of two points as a descriptor to construct the whole descriptors of a face. To reduce computation, we use Principle component analysis (PCA) to reduce the dimensions of descriptors. Moreover, either Support Vector Machines (SVMs) or K-Nearest-Neighbor (K-NN) is used to increase recognition rates. In contrast with the conventional recognition algorithm such as Eigenfaces, our method performs better under varying illumination because we use global shape features rather than gray scale pixel values. At last, we demonstrate our approach by experiments.