ERROR PROTECTION OF PCA BASED FACE RECOGNITION FOR TRANSMISSION OVER NOISY CHANNELS

Abstract

In this paper, we have proposed a system for reliable communication of coded grayscale facial images over noisy channels. Principle Component Analysis (PCA) is used for face dimensional reduction and face recognition and Repeated Bit Representation (RBR) is proposed to code the eigenfaces for mobile communication applications. Recognition obtained with RBR system approach reaches to %82 for 80 test images (2 poses per person) and 320 training images (8 poses per person) of ORL, by considering only the most important 30 coefficients of principle components in a noisy channel with SNR=-2dB