Cross correlation measure for decision fusion among multiple face classifiers

Abstract

We have developed a classifier decision fusion measure which is used as framework for combining multiple classifier decisions. The combination of different sources of information about a face, in the form of different feature sets and classification methods, provides an opportunity to develop an improved level of verification compared to the use of a single set of classifiers. Recently, the face recognition method based on Principal Component Analysis (PCA) and Directional Filter Bank (DFB) responses is integrated with voting algorithm. We look at the possibility of using cross correlation as a measure to compare the outputs of various classifiers. In our system recognition ability of the PCA is enhanced by providing directional images as inputs and then using the normalized cross correlation as a decision fusion measure. The proposed method fuses the decisions of DFB-PCA on the basis of maximum cross correlation of each directional test image with mean of its respective directional class. The experiment results showed the remarkable recognition rate of 97% in Olivetti data set.