A comprehensive experimental comparison of the aggregation techniques for face recognition

In face recognition, one of the most important problems to tackle is a large amount of data and the redundancy of information contained in facial images. There are numerous approaches attempting to reduce this redundancy. One of them is information aggregation based on the results of classifiers built on selected facial areas being the most salient regions from the point of view of classification both by humans and computers. In this study, we report on a series of experiments and offer a comprehensive comparison between various methods of aggregation of outputs of these classifiers based on essential facial features such as eyebrows, eyes, nose, and mouth areas. For each of them, we carry the recognition process utilizing the well-known Fisherfaces transformation. During the comparisons of the vectors representing the features of images (faces) after the transformations, we consider 16 similarity$/$dissimilarity measures for which we select the best aggregation operator. The set of operators to compare was selected on a basis of the comprehensive literature review regarding aggregation functions.