A new feature extraction based on advanced PCA for real time face recognition
Sara Mahdi, Mohammad Bagher Menhaj, Ali Mahdavi Hormat · 2013
Face recognition is an important and challenging task in machine vision. In this paper, we present a new method for feature extraction in the face recognition field. Firstly, each image in the training set is divided into a finite number of small parts. After that all parts of images merged together in one set, the K-means method is then applied over the set to obtain K clusters based on which a proper dictionary is constructed. High level features for face images are obtained by employing PCA on this dictionary. A robust face recognition system is developed based on multilayer perceptron neural networks (MLPs) and support vector machine (SVM) by imposing these features as the inputs. Experimental results easily show high accuracy of the system in terms of the correct recognition rate of 91% with low error rate and low computational complexity as well.