An Enhanced Two-Pass Classifier for Face Recognition Using Back-Propagation Neural Networks and Support Vector Machines

Abstract

A novel two-pass classification technique is presented in this paper to apply to the face recognition problem. Two approaches are integrated together to solve the face recognition problem efficiently: Back-propagation neural networks (BPNN) and Support Vector Machines (SVM). The face recognition task is realized in two passes. An intermediate result of classification is obtained in the first pass for an input using BPNN which indicates the classification status, and the second pass re-classifies the sample using SVM, considering the input sample and the intermediate result. Feature extraction is performed using principal component analysis (PCA) and the final classifier is designed by combining the advantages of artificial neural network and support vector machine. Testing on LFW and AT&T benchmark datasets showed that the proposed system outperforms state of the art face recognition techniques.

Key words: Back-propagation neural network, Face recognition, Principal Component Analysis, Support Vector Machine